It lets a third party verify a packaged computational claim offline, with one command, without access to the original environment.
I built it solo, after hours, while working construction, using AI tools heavily. I kept running into the same wall: even when a result looks good, there's no simple way for someone else to check it independently without re-running the full environment or trusting the number on faith.
That problem shows up everywhere:\n- ML: "our model reached 94.3% accuracy"\n- materials: "our simulation matches lab data within 1%"\n- pharma: "our pipeline passed quality checks"\n- finance: "our risk model was independently validated"
Different domains, same structure.
---
The gap
MLflow / W&B / DVC / Sigstore / SLSA solve adjacent problems well. What they don't provide is an offline third-party verification step with a semantic layer for the claim itself. File integrity alone is not enough.
The bypass attack:\n1. remove core semantic evidence (job_snapshot)\n2. recompute all SHA-256 hashes\n3. rebuild the manifest\n4. submit
A hash-only check still passes. MetaGenesis Core adds a second layer:\n- integrity layer \u2192 PASS\n- semantic layer \u2192 FAIL (job_snapshot missing)
That attack is an adversarial test in the public repo.
---
How it works
Layer 1 \u2014 integrity: SHA-256 per file + root hash\nLayer 2 \u2014 semantic: required fields present, payload.kind matches claim type, provenance intact
python scripts/mg.py verify --pack /path/to/bundle\n \u2192 PASS\n \u2192 FAIL: job_snapshot missing\n \u2192 FAIL: payload.kind does not match registered claim\n---
Current state
python scripts/steward_audit.py \u2192 PASS\n python -m pytest tests/ -q \u2192 91 passed\n python demos/open_data_demo_01/run_demo.py \u2192 PASS / PASS\n---
Honest limitations
Not validated by an external production team yet. The protocol works on the public codebase and tests, the adversarial scenario is caught, the demo is reproducible \u2014 but real-world integration still needs proof.
Limitations are machine-readable in reports/known_faults.yaml.
That first external "yes, this worked on our pipeline" is what I'm looking for.
---
If you think this is flawed, I want to know where. If it overlaps with an existing tool I'm missing, I want to know that too.
Site: https://metagenesis-core.dev\n Repo: https://github.com/Lama999901/metagenesis-core-public\n Contact: yehor@metagenesis-core.dev\n\n Inventor: Yehor Bazhynov\n Patent pending: USPTO #63/996,819It lets a third party verify a packaged computational claim offline, with one command, without access to the original environment.
I built it solo, after hours, while working construction, using AI tools heavily. I kept running into the same wall: even when a result looks good, there's no simple way for someone else to check it independently without re-running the full environment or trusting the number on faith.
That problem shows up everywhere:\n- ML: "our model reached 94.3% accuracy"\n- materials: "our simulation matches lab data within 1%"\n- pharma: "our pipeline passed quality checks"\n- finance: "our risk model was independently validated"
Different domains, same structure.
---
The gap
MLflow / W&B / DVC / Sigstore / SLSA solve adjacent problems well. What they don't provide is an offline third-party verification step with a semantic layer for the claim itself. File integrity alone is not enough.
The bypass attack:\n1. remove core semantic evidence (job_snapshot)\n2. recompute all SHA-256 hashes\n3. rebuild the manifest\n4. submit
A hash-only check still passes. MetaGenesis Core adds a second layer:\n- integrity layer \u2192 PASS\n- semantic layer \u2192 FAIL (job_snapshot missing)
That attack is an adversarial test in the public repo.
---
How it works
Layer 1 \u2014 integrity: SHA-256 per file + root hash\nLayer 2 \u2014 semantic: required fields present, payload.kind matches claim type, provenance intact
python scripts/mg.py verify --pack /path/to/bundle\n \u2192 PASS\n \u2192 FAIL: job_snapshot missing\n \u2192 FAIL: payload.kind does not match registered claim\n---
Current state
python scripts/steward_audit.py \u2192 PASS\n python -m pytest tests/ -q \u2192 91 passed\n python demos/open_data_demo_01/run_demo.py \u2192 PASS / PASS\n---
Honest limitations
Not validated by an external production team yet. The protocol works on the public codebase and tests, the adversarial scenario is caught, the demo is reproducible \u2014 but real-world integration still needs proof.
Limitations are machine-readable in reports/known_faults.yaml.
That first external "yes, this worked on our pipeline" is what I'm looking for.
---
If you think this is flawed, I want to know where. If it overlaps with an existing tool I'm missing, I want to know that too.
Site: https://metagenesis-core.dev\n Repo: https://github.com/Lama999901/metagenesis-core-public\n Contact: yehor@metagenesis-core.dev\n\n Inventor: Yehor Bazhynov\n Patent pending: USPTO #63/996,819* Battery-Included PostgreSQL Distribution, with PostGIS, TimescaleDB, Citus united as one.\n* Incredible observability powered by Prometheus & Grafana stack.\n* Self-healing HA PGSQL cluster, powered by patroni, haproxy, etcd\u2026\n* Auto-Configured PITR, powered by pgbackrest and optional MinIO cluster\n* Declarative API, Database-as-Code implemented with Ansible playbooks.\n* Versatile UseCases, Run Docker Apps, Run demos, Visualize data with ECharts panels.\n* Install in one command, provision IaaS with Terraform, and test with local Vagrant sandbox.
Features: https://github.com/Vonng/pigsty/blob/master/docs/FEATURE.md
I'm Vonng, the author of Pigsty and a PostgreSQL DBA of a large deployment (25K cores, 1PB TP data). Pigsty alleviates my pain of managing such a massive elephant in production.
It works well for us, and I'm glad to make it open-source. Since cloud RDS PostgreSQL are so expensive, sometimes it charges 10x compared to underlying EC2 and storage. An open-source alternative could help.
I'm happy to answer any questions about Pigsty ;)"},"title":{"matchLevel":"none","matchedWords":[],"value":"How about an open-source RDS alternative for PostgreSQL?"}},"_tags":["story","author_Vonng","story_34966165","ask_hn"],"author":"Vonng","created_at":"2023-02-28T05:59:00Z","created_at_i":1677563940,"num_comments":0,"objectID":"34966165","points":4,"story_id":34966165,"story_text":"Yo, Pigsty is a local-first open-source alternative for RDS PostgreSQL, with monitoring, HA, PITR, IaC, and lots of stuff to run a production-grade PostgreSQL Service on your nodes for free. Demo: http://demo.pigsty.cc.
* Battery-Included PostgreSQL Distribution, with PostGIS, TimescaleDB, Citus united as one.\n* Incredible observability powered by Prometheus & Grafana stack.\n* Self-healing HA PGSQL cluster, powered by patroni, haproxy, etcd\u2026\n* Auto-Configured PITR, powered by pgbackrest and optional MinIO cluster\n* Declarative API, Database-as-Code implemented with Ansible playbooks.\n* Versatile UseCases, Run Docker Apps, Run demos, Visualize data with ECharts panels.\n* Install in one command, provision IaaS with Terraform, and test with local Vagrant sandbox.
Features: https://github.com/Vonng/pigsty/blob/master/docs/FEATURE.md
I'm Vonng, the author of Pigsty and a PostgreSQL DBA of a large deployment (25K cores, 1PB TP data). Pigsty alleviates my pain of managing such a massive elephant in production.
It works well for us, and I'm glad to make it open-source. Since cloud RDS PostgreSQL are so expensive, sometimes it charges 10x compared to underlying EC2 and storage. An open-source alternative could help.
I'm happy to answer any questions about Pigsty ;)","title":"How about an open-source RDS alternative for PostgreSQL?","updated_at":"2026-01-12T18:44:33Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"whooocareslol"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I\u2019m less worried about being replaced by AI and more frustrated that companies are stealing our data to train AI models they profit from with potential to make us less valuable over time.
Whether you\u2019re:
- A coder writing clean, reusable functions or internal tooling,
- A UGC creator making tutorials or product demos,
- A data labeller doing precise annotations...
\u2026all of that labor creates intellectual property that ends up training AI models.
But here\u2019s the problem: we don\u2019t own any of it, even though it wouldn\u2019t exist without us.
They take our data\u2014by hook or by crook\u2014train a model, and extract massive value from it, while paying us nothing or, at best, a small one-time fee.
Yes, companies do play a valuable role. But they are using our work to replace us or devalue our work. So we have every right to ask for more.
If you really think about it, data mining is much like mineral mining \u2014 just as companies extract valuable resources like gold or diamonds from the earth, often exploiting labor and poorly governed regions, data mining extracts value from a poorly managed pool of people and their data, frequently without their full knowledge or consent regarding how it will be used.
I think now is the right time to build fairer systems around data for everyone\u2014royalties? data unions? open ownership of internal contributions within companies?
This business model isn't new\u2014some data sourcing and collection companies charge not only a one-time fee but also a usage-based fee each time the data is used.
Doing this is not only necessary to make the data supply chain fair, but also to improve AI. We all know that AI performance scales with compute, and the best way to leverage increasing compute is by applying it to new data. So, if we want AI to continue improving, we need a proper data supply chain. And if we want high-quality data for more complex tasks, we must ensure that everyone is paid fairly.
Would love to hear your thoughts on this."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"We give data to train AI models and get nothing in return"}},"_tags":["story","author_whooocareslol","story_44021995","ask_hn"],"author":"whooocareslol","children":[44022942,44023254],"created_at":"2025-05-18T15:16:27Z","created_at_i":1747581387,"num_comments":2,"objectID":"44021995","points":3,"story_id":44021995,"story_text":"I\u2019m less worried about being replaced by AI and more frustrated that companies are stealing our data to train AI models they profit from with potential to make us less valuable over time.
Whether you\u2019re:
- A coder writing clean, reusable functions or internal tooling,
- A UGC creator making tutorials or product demos,
- A data labeller doing precise annotations...
\u2026all of that labor creates intellectual property that ends up training AI models.
But here\u2019s the problem: we don\u2019t own any of it, even though it wouldn\u2019t exist without us.
They take our data\u2014by hook or by crook\u2014train a model, and extract massive value from it, while paying us nothing or, at best, a small one-time fee.
Yes, companies do play a valuable role. But they are using our work to replace us or devalue our work. So we have every right to ask for more.
If you really think about it, data mining is much like mineral mining \u2014 just as companies extract valuable resources like gold or diamonds from the earth, often exploiting labor and poorly governed regions, data mining extracts value from a poorly managed pool of people and their data, frequently without their full knowledge or consent regarding how it will be used.
I think now is the right time to build fairer systems around data for everyone\u2014royalties? data unions? open ownership of internal contributions within companies?
This business model isn't new\u2014some data sourcing and collection companies charge not only a one-time fee but also a usage-based fee each time the data is used.
Doing this is not only necessary to make the data supply chain fair, but also to improve AI. We all know that AI performance scales with compute, and the best way to leverage increasing compute is by applying it to new data. So, if we want AI to continue improving, we need a proper data supply chain. And if we want high-quality data for more complex tasks, we must ensure that everyone is paid fairly.
Would love to hear your thoughts on this.","title":"We give data to train AI models and get nothing in return","updated_at":"2025-05-19T06:44:29Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"h4l"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"v8serialize is a Python library that reads and writes V8's value serialization format[1][2], which is how V8 moves JS values between contexts when using postMessage(value), storing data in IndexedDB and Deno uses it to store values in its Deno KV database. Node.js and Deno have a v8.serialize() API to read/write the format. The format supports most of the conventional JavaScript types, like Map, Set, BigInt, Date, ArrayBuffer, Error, undefined etc; but not object prototypes or functions. It's a bit like JSON with more types and reference cycles.
The strength of the format is that it lets JavaScript programs serialize values without thinking about reducing types down to the subset that JSON supports. A weakness is that all the quirks of JavaScript's types need to be exposed by another language deserializing it, so it's certainly not ideal as a general-purpose format, just when ease of use from JavaScript is a priority.
This library lets Python programs work with JavaScript values by reading and writing this serialization format. As well as encoding/decoding the format itself, the library implements Python types that replicate the behaviour of JavaScript's Object, Array, Map, Set, etc, so that values can be round-tripped faithfully and quirks like JavaScript's sparse arrays don't cause Python to allocate massive contiguous lists.
I made this in an unplanned yak shave when setting out to write a client for Deno's KV database in Python. I'd assumed it was just storing JSON-encoded values, but I quickly realised it could round-trip all the regular JavaScript types, like Date & undefined, and realised it was using V8's value serialization format to store values at rest.
I'm now back to implementing a Deno KV client using this library. It can imagine it could also be used for other things, such as having a Python program act like a JavaScript Web Worker, receiving JavaScript values from postMessage()-like calls. Or perhaps could be used to preserve full JavaScript types in something using Python in the browser (in conjunction with a pure-js v8.serialize() implementation, like https://github.com/worker-tools/v8-value-serializer).
1: https://h4l.github.io/v8serialize/en/latest/explanation/v8_s...\n2: https://github.com/v8/v8/blob/main/src/objects/value-seriali..."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: V8serialize \u2013 Read/write V8-serialized JavaScript values from Python"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/h4l/v8serialize"}},"_tags":["story","author_h4l","story_41637623","show_hn"],"author":"h4l","children":[41637690],"created_at":"2024-09-24T15:31:38Z","created_at_i":1727191898,"num_comments":2,"objectID":"41637623","points":3,"story_id":41637623,"story_text":"v8serialize is a Python library that reads and writes V8's value serialization format[1][2], which is how V8 moves JS values between contexts when using postMessage(value), storing data in IndexedDB and Deno uses it to store values in its Deno KV database. Node.js and Deno have a v8.serialize() API to read/write the format. The format supports most of the conventional JavaScript types, like Map, Set, BigInt, Date, ArrayBuffer, Error, undefined etc; but not object prototypes or functions. It's a bit like JSON with more types and reference cycles.
The strength of the format is that it lets JavaScript programs serialize values without thinking about reducing types down to the subset that JSON supports. A weakness is that all the quirks of JavaScript's types need to be exposed by another language deserializing it, so it's certainly not ideal as a general-purpose format, just when ease of use from JavaScript is a priority.
This library lets Python programs work with JavaScript values by reading and writing this serialization format. As well as encoding/decoding the format itself, the library implements Python types that replicate the behaviour of JavaScript's Object, Array, Map, Set, etc, so that values can be round-tripped faithfully and quirks like JavaScript's sparse arrays don't cause Python to allocate massive contiguous lists.
I made this in an unplanned yak shave when setting out to write a client for Deno's KV database in Python. I'd assumed it was just storing JSON-encoded values, but I quickly realised it could round-trip all the regular JavaScript types, like Date & undefined, and realised it was using V8's value serialization format to store values at rest.
I'm now back to implementing a Deno KV client using this library. It can imagine it could also be used for other things, such as having a Python program act like a JavaScript Web Worker, receiving JavaScript values from postMessage()-like calls. Or perhaps could be used to preserve full JavaScript types in something using Python in the browser (in conjunction with a pure-js v8.serialize() implementation, like https://github.com/worker-tools/v8-value-serializer).
1: https://h4l.github.io/v8serialize/en/latest/explanation/v8_s...\n2: https://github.com/v8/v8/blob/main/src/objects/value-seriali...","title":"Show HN: V8serialize \u2013 Read/write V8-serialized JavaScript values from Python","updated_at":"2024-09-25T02:01:17Z","url":"https://github.com/h4l/v8serialize"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"harperlabs"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"We've been running reliability audits on AI agents before production deployment, and the failure patterns are consistent enough that I built a framework around them.
Some context on why this matters right now: Gartner predicted over 40% of AI agent projects will fail by 2027. In January 2026, a prompt injection in a customer support agent processed a $47,000 fraudulent refund. These aren't fringe cases anymore.
The 7 failure modes we see most often:
1. Hallucination under unexpected inputs \u2014 works perfectly in demos, invents data when the input is slightly off
2. Edge case collapse \u2014 null values, Unicode names (O'Brien, Jos\u00e9, \u5317\u4eac), empty fields, concurrent requests
3. Prompt injection \u2014 if your agent processes external content, users can hijack its behavior through that content
4. Context limit surprises \u2014 agent works for 95% of conversations, then silently misbehaves when the context window fills. No error. Just wrong behavior.
5. Cascade failures \u2014 tool call #1 fails, agent keeps going, by the time a human sees the result 3 calls have compounded the error
6. Data integration drift \u2014 built against your schema in January, schema changed in February, still calling deprecated endpoints in March
7. Authorization confusion \u2014 multi-tenant system, cached context from User A bleeds into User B's session
We've built 50+ test cases across these categories. Most teams test #1 and #3. Almost no one systematically tests #4, #5, and #6 before shipping.
Happy to share the framework. Curious what failure modes you've hit that I haven't listed."},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: How are you testing AI agents before shipping to production?"}},"_tags":["story","author_harperlabs","story_47325105","ask_hn"],"author":"harperlabs","children":[47325268,47326204,47328379,47371804,47420654,47424704],"created_at":"2026-03-10T16:04:07Z","created_at_i":1773158647,"num_comments":6,"objectID":"47325105","points":2,"story_id":47325105,"story_text":"We've been running reliability audits on AI agents before production deployment, and the failure patterns are consistent enough that I built a framework around them.
Some context on why this matters right now: Gartner predicted over 40% of AI agent projects will fail by 2027. In January 2026, a prompt injection in a customer support agent processed a $47,000 fraudulent refund. These aren't fringe cases anymore.
The 7 failure modes we see most often:
1. Hallucination under unexpected inputs \u2014 works perfectly in demos, invents data when the input is slightly off
2. Edge case collapse \u2014 null values, Unicode names (O'Brien, Jos\u00e9, \u5317\u4eac), empty fields, concurrent requests
3. Prompt injection \u2014 if your agent processes external content, users can hijack its behavior through that content
4. Context limit surprises \u2014 agent works for 95% of conversations, then silently misbehaves when the context window fills. No error. Just wrong behavior.
5. Cascade failures \u2014 tool call #1 fails, agent keeps going, by the time a human sees the result 3 calls have compounded the error
6. Data integration drift \u2014 built against your schema in January, schema changed in February, still calling deprecated endpoints in March
7. Authorization confusion \u2014 multi-tenant system, cached context from User A bleeds into User B's session
We've built 50+ test cases across these categories. Most teams test #1 and #3. Almost no one systematically tests #4, #5, and #6 before shipping.
Happy to share the framework. Curious what failure modes you've hit that I haven't listed.","title":"Ask HN: How are you testing AI agents before shipping to production?","updated_at":"2026-03-23T01:03:04Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"glaugh"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Data from Yelp's Dataset Challenge"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["delos"],"value":"https://www.statwing.com/demos/yelp-dataset-challenge#workspaces/21456"}},"_tags":["story","author_glaugh","story_8238759"],"author":"glaugh","children":[8238822,8238943,8239309],"created_at":"2014-08-28T18:05:58Z","created_at_i":1409249158,"num_comments":6,"objectID":"8238759","points":56,"story_id":8238759,"story_text":"","title":"Data from Yelp's Dataset Challenge","updated_at":"2024-09-19T21:06:26Z","url":"https://www.statwing.com/demos/yelp-dataset-challenge#workspaces/21456"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"NimaBoscarino"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hugging Face Hub: discover and share ML models, datasets, and demos"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://huggingface.co/docs/hub/index"}},"_tags":["story","author_NimaBoscarino","story_31586923"],"author":"NimaBoscarino","created_at":"2022-06-01T19:40:12Z","created_at_i":1654112412,"num_comments":0,"objectID":"31586923","points":2,"story_id":31586923,"title":"Hugging Face Hub: discover and share ML models, datasets, and demos","updated_at":"2024-09-20T11:19:01Z","url":"https://huggingface.co/docs/hub/index"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mrlou"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Demohub.dev Democratizing Access to Demos for Modern Data Teams"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.demohub.dev"}},"_tags":["story","author_mrlou","story_31213100"],"author":"mrlou","created_at":"2022-04-30T03:27:37Z","created_at_i":1651289257,"num_comments":0,"objectID":"31213100","points":2,"story_id":31213100,"title":"Demohub.dev Democratizing Access to Demos for Modern Data Teams","updated_at":"2024-09-20T11:00:11Z","url":"https://www.demohub.dev"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"1SockChuck"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"First Look: Dell's New Modular Data Center"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"http://www.datacenterknowledge.com/archives/2010/09/09/first-look-dells-new-modular-data-center/"}},"_tags":["story","author_1SockChuck","story_1675250"],"author":"1SockChuck","created_at":"2010-09-09T13:23:35Z","created_at_i":1284038615,"num_comments":0,"objectID":"1675250","points":1,"story_id":1675250,"story_text":"","title":"First Look: Dell's New Modular Data Center","updated_at":"2024-09-19T17:14:25Z","url":"http://www.datacenterknowledge.com/archives/2010/09/09/first-look-dells-new-modular-data-center/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"chenyuanjen"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"[Event] Real-Time Analytics to Stop Data Breach: Demos and Discussion"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"http://www.meetup.com/Palo-Alto-Data-Science-Association/events/224473199/"}},"_tags":["story","author_chenyuanjen","story_10055964"],"author":"chenyuanjen","created_at":"2015-08-13T17:50:15Z","created_at_i":1439488215,"num_comments":0,"objectID":"10055964","points":1,"story_id":10055964,"story_text":"","title":"[Event] Real-Time Analytics to Stop Data Breach: Demos and Discussion","updated_at":"2024-09-19T22:12:32Z","url":"http://www.meetup.com/Palo-Alto-Data-Science-Association/events/224473199/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"twwch"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hi everyone,
I've been working on AI Agents recently and noticed a common frustration: most AI diagramming tools generate static images (PNG/JPG). If the AI gets one detail wrong, you can't easily fix it\u2014you have to regenerate the whole thing.
I wanted a tool that generates editable code and renders it interactively. So I built DeepDiagram.
It\u2019s an open-source web platform that uses a Multi-Agent architecture. Instead of a single prompt trying to do everything, a Router analyzes your intent and directs it to a specialized agent (e.g., a Draw.io expert, a Data Visualization expert, or a Flowchart expert).
Demos\nMind Map & Data Charts (ECharts):
Draw.io Generation (XML based):
Mermaid Diagrams:
Tech Stack\nOrchestration: LangGraph (Python) for the ReAct loop and state management.
Frontend: React 19, Vite, TailwindCSS, Zustand.
Backend: FastAPI (Python 3.10+).
Streaming: Uses SSE (Server-Sent Events) for real-time preview (typewriter effect).
Rendering Engines: React Flow, Apache ECharts, Mermaid.js, and Draw.io (Atlas theme).
Key Features\nDraw.io Agent: Generates compatible XML. You can edit the result directly in the canvas.
Data Chart Agent: Paste raw data (CSV/JSON/Text), and it visualizes it using ECharts (Bar/Line/Pie).
Branching & Time Travel: Since AI isn't perfect, I implemented a Git-like branching system. You can "retry" a message, creating a new branch, or navigate back to previous states to see the agent's thought process.
Self-Hosted: Docker Compose support included. Works great with OpenAI or DeepSeek V3 (highly recommended for cost/performance).
Links\nGitHub (AGPL-3.0): https://github.com/twwch/DeepDiagram
It's still in the early stages, but I'd love to hear your feedback or see your PRs!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Open source AI diagram generator (Draw.io, Mermaid)"}},"_tags":["story","author_twwch","story_46487082","ask_hn"],"author":"twwch","created_at":"2026-01-04T11:47:36Z","created_at_i":1767527256,"num_comments":0,"objectID":"46487082","points":2,"story_id":46487082,"story_text":"Hi everyone,
I've been working on AI Agents recently and noticed a common frustration: most AI diagramming tools generate static images (PNG/JPG). If the AI gets one detail wrong, you can't easily fix it\u2014you have to regenerate the whole thing.
I wanted a tool that generates editable code and renders it interactively. So I built DeepDiagram.
It\u2019s an open-source web platform that uses a Multi-Agent architecture. Instead of a single prompt trying to do everything, a Router analyzes your intent and directs it to a specialized agent (e.g., a Draw.io expert, a Data Visualization expert, or a Flowchart expert).
Demos\nMind Map & Data Charts (ECharts):
Draw.io Generation (XML based):
Mermaid Diagrams:
Tech Stack\nOrchestration: LangGraph (Python) for the ReAct loop and state management.
Frontend: React 19, Vite, TailwindCSS, Zustand.
Backend: FastAPI (Python 3.10+).
Streaming: Uses SSE (Server-Sent Events) for real-time preview (typewriter effect).
Rendering Engines: React Flow, Apache ECharts, Mermaid.js, and Draw.io (Atlas theme).
Key Features\nDraw.io Agent: Generates compatible XML. You can edit the result directly in the canvas.
Data Chart Agent: Paste raw data (CSV/JSON/Text), and it visualizes it using ECharts (Bar/Line/Pie).
Branching & Time Travel: Since AI isn't perfect, I implemented a Git-like branching system. You can "retry" a message, creating a new branch, or navigate back to previous states to see the agent's thought process.
Self-Hosted: Docker Compose support included. Works great with OpenAI or DeepSeek V3 (highly recommended for cost/performance).
Links\nGitHub (AGPL-3.0): https://github.com/twwch/DeepDiagram
It's still in the early stages, but I'd love to hear your feedback or see your PRs!","title":"Open source AI diagram generator (Draw.io, Mermaid)","updated_at":"2026-03-05T23:19:16Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"martypitt"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hey HN,
I'm Marty, founder of Orbital[0]\u2014a semantic data integration platform.
As we've been demo'ing Orbital to people, we kept hitting a common challenge: when people evaluate Orbital, they need a demo to play with. But since Orbital is all about integration, those demos need real data sources\u2014Kafka topics, S3 buckets, SQL databases, microservices, etc.\u2014that actually do stuff: emit messages, respond to API calls, and contain data.
Historically, we shipped demos as a mix of Docker Compose files and Spring Boot microservices (for the "doing stuff" part). But that was painful\u2014Spring Boot services had to be dockerized and deployed somewhere, and code got messy.
So we built Nebula[1] to fix this. It\u2019s a simple DSL where you declare the Docker images you need, then write Kotlin code to script them\u2014all in a single file.
Nebula itself runs as a standalone Docker image. At runtime, it takes in the script, pulls and starts the required Docker images, then executes the Kotlin code you've written. The docker parts are handled by the awesome TestContainers, and scripting by Kotlin. Since it's just Kotlin, you can write whatever you need\u2014we've even simulated full food delivery systems inside Nebula scripts.
We split it into its own repo under Apache 2. It's simple but surprisingly powerful (and kinda fun) for quickly hacking together demo ecosystems.
Thought I'd share!
[0]: Orbital - https://github.com/orbitalapi/orbital
[1]: Nebula - https://github.com/orbitalapi/nebula"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["delos"],"value":"Show HN: Nebula \u2013 A DSL for scripting TestContainers-based demos"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/orbitalapi/nebula"}},"_tags":["story","author_martypitt","story_43253466","show_hn"],"author":"martypitt","created_at":"2025-03-04T12:00:34Z","created_at_i":1741089634,"num_comments":0,"objectID":"43253466","points":2,"story_id":43253466,"story_text":"Hey HN,
I'm Marty, founder of Orbital[0]\u2014a semantic data integration platform.
As we've been demo'ing Orbital to people, we kept hitting a common challenge: when people evaluate Orbital, they need a demo to play with. But since Orbital is all about integration, those demos need real data sources\u2014Kafka topics, S3 buckets, SQL databases, microservices, etc.\u2014that actually do stuff: emit messages, respond to API calls, and contain data.
Historically, we shipped demos as a mix of Docker Compose files and Spring Boot microservices (for the "doing stuff" part). But that was painful\u2014Spring Boot services had to be dockerized and deployed somewhere, and code got messy.
So we built Nebula[1] to fix this. It\u2019s a simple DSL where you declare the Docker images you need, then write Kotlin code to script them\u2014all in a single file.
Nebula itself runs as a standalone Docker image. At runtime, it takes in the script, pulls and starts the required Docker images, then executes the Kotlin code you've written. The docker parts are handled by the awesome TestContainers, and scripting by Kotlin. Since it's just Kotlin, you can write whatever you need\u2014we've even simulated full food delivery systems inside Nebula scripts.
We split it into its own repo under Apache 2. It's simple but surprisingly powerful (and kinda fun) for quickly hacking together demo ecosystems.
Thought I'd share!
[0]: Orbital - https://github.com/orbitalapi/orbital
[1]: Nebula - https://github.com/orbitalapi/nebula","title":"Show HN: Nebula \u2013 A DSL for scripting TestContainers-based demos","updated_at":"2025-03-04T13:26:15Z","url":"https://github.com/orbitalapi/nebula"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"benja123"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"WorldFixture gives you emulated third-party services, object storage, and email, all backed by the same fake company so the data and API responses feel real.
So, for example, if your app uses Stripe, Gmail, Slack, and GitHub, you can call the same APIs through WorldFixture and get responses containing data from the fake company. The data is organized so it tells a coherent story across the different services, making the test/demo feel real. Moreover, you can stream events over time so the app feels "alive."
It is useful for integration and E2E testing, as well as interactive demos.
It was originally part of droplive.io, which lets you try hosted open-source software in a time-boxed microVM. I realized that demos with no data kind of suck, so I started adding data and emulated third-party services, which eventually became WorldFixture.
You can also create your own datasets/worlds if the default fake company is not what you need."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Worldfixture \u2013 An OSS fake company your software can interact with"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://worldfixture.com/"}},"_tags":["story","author_benja123","story_49603266","show_hn"],"author":"benja123","created_at":"2026-09-07T21:42:13Z","created_at_i":1788817333,"num_comments":0,"objectID":"49603266","points":1,"story_id":49603266,"story_text":"WorldFixture gives you emulated third-party services, object storage, and email, all backed by the same fake company so the data and API responses feel real.
So, for example, if your app uses Stripe, Gmail, Slack, and GitHub, you can call the same APIs through WorldFixture and get responses containing data from the fake company. The data is organized so it tells a coherent story across the different services, making the test/demo feel real. Moreover, you can stream events over time so the app feels "alive."
It is useful for integration and E2E testing, as well as interactive demos.
It was originally part of droplive.io, which lets you try hosted open-source software in a time-boxed microVM. I realized that demos with no data kind of suck, so I started adding data and emulated third-party services, which eventually became WorldFixture.
You can also create your own datasets/worlds if the default fake company is not what you need.","title":"Show HN: Worldfixture \u2013 An OSS fake company your software can interact with","updated_at":"2026-09-07T21:47:45Z","url":"https://worldfixture.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ashishbagri"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I've built GlassGen to solve the common problem of generating real time synthetic data for testing, demos, and ML datasets. While Faker is great for individual data points, GlassGen adds:
- Configurable data publishing (CSV, Kafka, Webhooks)
- Precise rate control (records/second)
- Controlled data duplication
- Extensible architecture for custom generators and sinks
Key features:
- Built on top of Faker for reliable data generation
- Simple JSON/YAML configuration
- Support for complex data relationships
- Real-time data streaming to Kafka
- Custom sink implementations
GitHub: https://github.com/glassflow/glassgen
Docs: https://glassgen.glassflow.dev/
Would love feedback from the community, especially on:
1. Additional sink types that would be useful
2. Performance optimization opportunities
3. Ideas for handling more complex data relationships"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: A real world streaming data generator in Python"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/glassflow/glassgen"}},"_tags":["story","author_ashishbagri","story_43934537","show_hn"],"author":"ashishbagri","created_at":"2025-05-09T07:21:48Z","created_at_i":1746775308,"num_comments":0,"objectID":"43934537","points":1,"story_id":43934537,"story_text":"I've built GlassGen to solve the common problem of generating real time synthetic data for testing, demos, and ML datasets. While Faker is great for individual data points, GlassGen adds:
- Configurable data publishing (CSV, Kafka, Webhooks)
- Precise rate control (records/second)
- Controlled data duplication
- Extensible architecture for custom generators and sinks
Key features:
- Built on top of Faker for reliable data generation
- Simple JSON/YAML configuration
- Support for complex data relationships
- Real-time data streaming to Kafka
- Custom sink implementations
GitHub: https://github.com/glassflow/glassgen
Docs: https://glassgen.glassflow.dev/
Would love feedback from the community, especially on:
1. Additional sink types that would be useful
2. Performance optimization opportunities
3. Ideas for handling more complex data relationships","title":"Show HN: A real world streaming data generator in Python","updated_at":"2025-05-09T09:00:00Z","url":"https://github.com/glassflow/glassgen"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"edichief"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I've built a tool to create multimodal datasets for demos and PoC with fully customizable data fields and an interface to evaluate data quality!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: Generate Multimodal Synthetic Data"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.gen9000.co/"}},"_tags":["story","author_edichief","story_39885351","show_hn"],"author":"edichief","created_at":"2024-03-31T15:53:50Z","created_at_i":1711900430,"num_comments":0,"objectID":"39885351","points":1,"story_id":39885351,"story_text":"I've built a tool to create multimodal datasets for demos and PoC with fully customizable data fields and an interface to evaluate data quality!","title":"Show HN: Generate Multimodal Synthetic Data","updated_at":"2024-09-20T16:45:50Z","url":"https://www.gen9000.co/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sprokuski"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Automating highly targeted demos: Visualizing free API data for better leads"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"https://reflect.io/blog/api-data/"}},"_tags":["story","author_sprokuski","story_14753723"],"author":"sprokuski","created_at":"2017-07-12T16:10:49Z","created_at_i":1499875849,"num_comments":0,"objectID":"14753723","points":9,"story_id":14753723,"title":"Automating highly targeted demos: Visualizing free API data for better leads","updated_at":"2024-09-20T01:04:18Z","url":"https://reflect.io/blog/api-data/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"dima55"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Shell data processing: vnlog demos"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["delos"],"value":"http://notes.secretsauce.net/notes/2018/05/21_more-vnlog-demos.html"}},"_tags":["story","author_dima55","story_30676859"],"author":"dima55","created_at":"2022-03-14T19:15:34Z","created_at_i":1647285334,"num_comments":0,"objectID":"30676859","points":6,"story_id":30676859,"title":"Shell data processing: vnlog demos","updated_at":"2024-09-20T10:45:13Z","url":"http://notes.secretsauce.net/notes/2018/05/21_more-vnlog-demos.html"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kbatdorf"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Collaborative beat machine demo using Deno's KV distributed cloud database"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://beats.deno.dev"}},"_tags":["story","author_kbatdorf","story_36371621"],"author":"kbatdorf","children":[36371622],"created_at":"2023-06-17T16:22:30Z","created_at_i":1687018950,"num_comments":1,"objectID":"36371621","points":2,"story_id":36371621,"title":"Collaborative beat machine demo using Deno's KV distributed cloud database","updated_at":"2024-09-20T14:26:34Z","url":"https://beats.deno.dev"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"nreece"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Intel demos 48-core cloud-datacenter-on-a-chip"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"http://arstechnica.com/business/news/2009/12/intel-demos-48-core-cloud-datacenter-on-a-chip.ars"}},"_tags":["story","author_nreece","story_979000"],"author":"nreece","created_at":"2009-12-06T02:06:28Z","created_at_i":1260065188,"num_comments":0,"objectID":"979000","points":1,"story_id":979000,"title":"Intel demos 48-core cloud-datacenter-on-a-chip","updated_at":"2024-09-19T16:52:08Z","url":"http://arstechnica.com/business/news/2009/12/intel-demos-48-core-cloud-datacenter-on-a-chip.ars"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"robbiejs"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I'm Robbert, the creator of DataGridXL.js. Last month I released version 2 which includes many new features.
DataGridXL is a free (and commercial) editable data table library written in ES6.
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.\n- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.\n- It has its own Virtual DOM implementation to prevent DOM errors.\n- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there. And let us know if you have any suggestions.
Please let me know if you have any suggestions or comments!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: Datagridxl2.js \u2013 Fast Excel-like data table library"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"https://www.datagridxl.com/"}},"_tags":["story","author_robbiejs","story_30919257","show_hn"],"author":"robbiejs","children":[30919771,30919869,30919914,30919946,30919986,30920078,30920103,30920112,30920419,30920492,30920563,30921093,30921345,30921441,30921493,30921565,30921600,30921833,30923759,30924097,30924386,30924761,30925790,30928611,30928774,30929279,30930007,30931083,30931881,30973190],"created_at":"2022-04-05T14:16:33Z","created_at_i":1649168193,"num_comments":104,"objectID":"30919257","points":301,"story_id":30919257,"story_text":"I'm Robbert, the creator of DataGridXL.js. Last month I released version 2 which includes many new features.
DataGridXL is a free (and commercial) editable data table library written in ES6.
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.\n- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.\n- It has its own Virtual DOM implementation to prevent DOM errors.\n- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there. And let us know if you have any suggestions.
Please let me know if you have any suggestions or comments!","title":"Show HN: Datagridxl2.js \u2013 Fast Excel-like data table library","updated_at":"2025-09-12T04:56:08Z","url":"https://www.datagridxl.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"StellarGraph"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"StellarGraph is an open-source library implementing a variety of state-of-the-art graph machine learning algorithms. The project is delivered as part of CSIRO\u2019s Data61.
We are happy to announce the 0.8.1 release of the library, which extends StellarGraph capability by adding new algorithms and demos, enhancing interpretability via saliency maps for Graph Attention (GAT), and further simplifying graph machine learning workflows through standardised model APIs and arguments.
This release, we\u2019ve dealt with some bugs from the previous release and introduced new features and enhancements. Some of these include:
- New directed GraphSAGE algorithm (a generalisation of GraphSAGE to directed graphs)\n- New Attri2vec algorithm \n- New PPNP and APPNP algorithms \n- New Graph Attention (GAT) saliency maps for interpreting node classification with Graph Attention Networks\n- Added directed SampledBFS walks on directed graphs\n- Unified API of GCN, GAT, GraphSAGE, and HinSAGE classes by adding build() method to GCN and GAT classes\n- Enhanced unsupervised GraphSage speed up via multithreading\n- Support of sparse generators in the GCN saliency map implementation.\n- Unified activations and regularisation for GraphSAGE, HinSAGE, GCN and GAT\n- Changed from using keras to tensorflow.keras
We\u2019ve also added new demos using real-world datasets to show how StellarGraph can solve these tasks.
Access the StellarGraph project and explore the new features on GitHub (https://github.com/stellargraph/stellargraph). StellarGraph is a Python 3 library.
We welcome your feedback and contributions.
With thanks, the StellarGraph team."},"title":{"matchLevel":"none","matchedWords":[],"value":"StellarGraph v0.8.1 Open-Source Python Machine Learning Library for Graphs"}},"_tags":["story","author_StellarGraph","story_21470801","ask_hn"],"author":"StellarGraph","created_at":"2019-11-07T05:57:47Z","created_at_i":1573106267,"num_comments":0,"objectID":"21470801","points":5,"story_id":21470801,"story_text":"StellarGraph is an open-source library implementing a variety of state-of-the-art graph machine learning algorithms. The project is delivered as part of CSIRO\u2019s Data61.
We are happy to announce the 0.8.1 release of the library, which extends StellarGraph capability by adding new algorithms and demos, enhancing interpretability via saliency maps for Graph Attention (GAT), and further simplifying graph machine learning workflows through standardised model APIs and arguments.
This release, we\u2019ve dealt with some bugs from the previous release and introduced new features and enhancements. Some of these include:
- New directed GraphSAGE algorithm (a generalisation of GraphSAGE to directed graphs)\n- New Attri2vec algorithm \n- New PPNP and APPNP algorithms \n- New Graph Attention (GAT) saliency maps for interpreting node classification with Graph Attention Networks\n- Added directed SampledBFS walks on directed graphs\n- Unified API of GCN, GAT, GraphSAGE, and HinSAGE classes by adding build() method to GCN and GAT classes\n- Enhanced unsupervised GraphSage speed up via multithreading\n- Support of sparse generators in the GCN saliency map implementation.\n- Unified activations and regularisation for GraphSAGE, HinSAGE, GCN and GAT\n- Changed from using keras to tensorflow.keras
We\u2019ve also added new demos using real-world datasets to show how StellarGraph can solve these tasks.
Access the StellarGraph project and explore the new features on GitHub (https://github.com/stellargraph/stellargraph). StellarGraph is a Python 3 library.
We welcome your feedback and contributions.
With thanks, the StellarGraph team.","title":"StellarGraph v0.8.1 Open-Source Python Machine Learning Library for Graphs","updated_at":"2024-09-20T05:12:21Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"robbiejs"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hello HN,
I\u2019m Robbert, the creator of DataGridXL.js. After 1,5 years of hard work I have just released version 2.
DataGridXL is a free (and commercial) editable data table library written in ES6.
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.
- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.
- It has its own Virtual DOM implementation to prevent DOM errors.
- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there. And let us know if you have any suggestions.
What's new in version 2:
* Rapidly search cells via top bar, just like Google Sheets.
* Freeze any amount of rows and/or columns.
* Hide and unhide rows and/or columns, just like in Google Sheets.
* Multi-Range Selection
* Fullscreen Mode
* Input bar at the bottom, so it works on touch devices too.
DataGridXL on Github: https://github.com/DataGridXL/DataGridXL2\nDataGridXL on NPM: https://www.npmjs.com/package/@datagridxl/datagridxl2\nDataGridXL Docs: https://datagridxl.com/docs
Please let me know if you have any suggestions or comments!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: Datagridxl.js v2 \u2013 no-nonsense fast Excel-like data table lib"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"https://www.datagridxl.com/"}},"_tags":["story","author_robbiejs","story_30491642","show_hn"],"author":"robbiejs","children":[30491694],"created_at":"2022-02-27T19:27:11Z","created_at_i":1645990031,"num_comments":1,"objectID":"30491642","points":4,"story_id":30491642,"story_text":"Hello HN,
I\u2019m Robbert, the creator of DataGridXL.js. After 1,5 years of hard work I have just released version 2.
DataGridXL is a free (and commercial) editable data table library written in ES6.
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.
- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.
- It has its own Virtual DOM implementation to prevent DOM errors.
- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there. And let us know if you have any suggestions.
What's new in version 2:
* Rapidly search cells via top bar, just like Google Sheets.
* Freeze any amount of rows and/or columns.
* Hide and unhide rows and/or columns, just like in Google Sheets.
* Multi-Range Selection
* Fullscreen Mode
* Input bar at the bottom, so it works on touch devices too.
DataGridXL on Github: https://github.com/DataGridXL/DataGridXL2\nDataGridXL on NPM: https://www.npmjs.com/package/@datagridxl/datagridxl2\nDataGridXL Docs: https://datagridxl.com/docs
Please let me know if you have any suggestions or comments!","title":"Show HN: Datagridxl.js v2 \u2013 no-nonsense fast Excel-like data table lib","updated_at":"2024-09-20T10:40:01Z","url":"https://www.datagridxl.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"BlueHotDog2"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hey all, we're in the process of scaling our sales org. Since we rely on demos we're exploring the best practices in managing demo environments. \nOur current thoughts are:\n- We can simply use production, but that seems risky - New features might surprise AEs in the demos, A/B tests, production data? data reset between demos\n- We could clone a branch of production and create a dedicated demo env, almost the same problems as above minus the features surprises.\n- We can simply use PowerPoint which solves most of the problems above but feels like we'll lose a lot of the wow factor of our app.
Any ideas? \nWhat are the current best practices to create/manage/update a demo environment?\nShare your wisdom guys!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: How do you manage your demo env?"}},"_tags":["story","author_BlueHotDog2","story_22576493","ask_hn"],"author":"BlueHotDog2","children":[22576622],"created_at":"2020-03-14T16:28:35Z","created_at_i":1584203315,"num_comments":2,"objectID":"22576493","points":3,"story_id":22576493,"story_text":"Hey all, we're in the process of scaling our sales org. Since we rely on demos we're exploring the best practices in managing demo environments. \nOur current thoughts are:\n- We can simply use production, but that seems risky - New features might surprise AEs in the demos, A/B tests, production data? data reset between demos\n- We could clone a branch of production and create a dedicated demo env, almost the same problems as above minus the features surprises.\n- We can simply use PowerPoint which solves most of the problems above but feels like we'll lose a lot of the wow factor of our app.
Any ideas? \nWhat are the current best practices to create/manage/update a demo environment?\nShare your wisdom guys!","title":"Ask HN: How do you manage your demo env?","updated_at":"2024-09-20T05:54:21Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jarichb"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hi HN,
I\u2019ve recently open-sourced model2data, a small tool that grew out of a recurring problem in my day-to-day work: needing good synthetic data for testing, demos, and development.
Random generators and faker-style approaches often produce data that looks plausible at first glance, but quickly breaks down once you care about relationships, constraints, or model semantics. model2data tries to stay closer to the data model itself, so the generated data remains structurally meaningful.
There\u2019s no big roadmap yet \u2014 it\u2019s something I\u2019ve found useful myself and wanted to share.
Repo: https://github.com/JB-Analytica/model2data
I\u2019d be very interested to hear how others approach synthetic data, or where this kind of approach might fall short."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: Model2data \u2013 generate realistic synthetic data from data models"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/JB-Analytica/model2data"}},"_tags":["story","author_jarichb","story_46496731","show_hn"],"author":"jarichb","children":[46496759],"created_at":"2026-01-05T09:09:30Z","created_at_i":1767604170,"num_comments":0,"objectID":"46496731","points":2,"story_id":46496731,"story_text":"Hi HN,
I\u2019ve recently open-sourced model2data, a small tool that grew out of a recurring problem in my day-to-day work: needing good synthetic data for testing, demos, and development.
Random generators and faker-style approaches often produce data that looks plausible at first glance, but quickly breaks down once you care about relationships, constraints, or model semantics. model2data tries to stay closer to the data model itself, so the generated data remains structurally meaningful.
There\u2019s no big roadmap yet \u2014 it\u2019s something I\u2019ve found useful myself and wanted to share.
Repo: https://github.com/JB-Analytica/model2data
I\u2019d be very interested to hear how others approach synthetic data, or where this kind of approach might fall short.","title":"Show HN: Model2data \u2013 generate realistic synthetic data from data models","updated_at":"2026-03-05T23:19:49Z","url":"https://github.com/JB-Analytica/model2data"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ChicagoBoy11"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I remember many years ago seeing a demo from an algorithm from MIT which took a simple raw video of a baby and, with clever video manipulation, could so clearly augment changes between frames that it made it even trivial for the naked eye to see the baby's pulse, or breathing rate (and it had accompanying demos which showed that data simply from the inferred image).
My extended family suffered an incredible tragedy many years ago after a cousin briefly asfixiated during sleep, causing irreversible brain damage, and it has constantly nagged me that surely some kind of baby monitoring leveraging that tech could have prevented it. As a newly-expectant father, I am turning my attention now to these questions for my own child, but as hard as I've looked, it seems to me that the technology behind baby monitoring hasn't strayed too far from what we've grown accustomed to for a long time, save for just improvements in ergonomics and things of the sort. Curious if anyone was aware of some examples of what today would be the most extensive types of monitors available, eyeing especially any leveraging of "tech" to infer things like oxigenation, pulse, etc."},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: AI Baby monitoring state of the art?"}},"_tags":["story","author_ChicagoBoy11","story_39538212","ask_hn"],"author":"ChicagoBoy11","created_at":"2024-02-28T14:16:40Z","created_at_i":1709129800,"num_comments":0,"objectID":"39538212","points":2,"story_id":39538212,"story_text":"I remember many years ago seeing a demo from an algorithm from MIT which took a simple raw video of a baby and, with clever video manipulation, could so clearly augment changes between frames that it made it even trivial for the naked eye to see the baby's pulse, or breathing rate (and it had accompanying demos which showed that data simply from the inferred image).
My extended family suffered an incredible tragedy many years ago after a cousin briefly asfixiated during sleep, causing irreversible brain damage, and it has constantly nagged me that surely some kind of baby monitoring leveraging that tech could have prevented it. As a newly-expectant father, I am turning my attention now to these questions for my own child, but as hard as I've looked, it seems to me that the technology behind baby monitoring hasn't strayed too far from what we've grown accustomed to for a long time, save for just improvements in ergonomics and things of the sort. Curious if anyone was aware of some examples of what today would be the most extensive types of monitors available, eyeing especially any leveraging of "tech" to infer things like oxigenation, pulse, etc.","title":"Ask HN: AI Baby monitoring state of the art?","updated_at":"2024-09-20T16:28:30Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"robbiejs"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hello HN,
I\u2019m Robbert, the creator of DataGridXL.js. After 1,5 years of hard work I have just released version 2.
DataGridXL is a free (and commercial) editable data table library written in ES6.
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.\n- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.\n- It has its own Virtual DOM implementation to prevent DOM errors.\n- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there. And let us know if you have any suggestions.
What's new in version 2:
* Rapidly search cells via top bar, just like Google Sheets.\n* Freeze any amount of rows and/or columns.\n* Hide and unhide rows and/or columns, just like in Google Sheets.\n* Multi-Range Selection\n* Fullscreen Mode\n* Input bar at the bottom, so it works on touch devices too.
DataGridXL on Github: https://github.com/DataGridXL/DataGridXL2\nDataGridXL on NPM: https://www.npmjs.com/package/@datagridxl/datagridxl2\nDataGridXL Docs: https://datagridxl.com/docs
Please let me know if you have any suggestions or comments!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: Datagridxl2.js \u2013 No-nonsense fast Excel-like data table library"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"https://www.datagridxl.com/"}},"_tags":["story","author_robbiejs","story_30444994","show_hn"],"author":"robbiejs","children":[30484416],"created_at":"2022-02-23T19:12:31Z","created_at_i":1645643551,"num_comments":4,"objectID":"30444994","points":1,"story_id":30444994,"story_text":"Hello HN,
I\u2019m Robbert, the creator of DataGridXL.js. After 1,5 years of hard work I have just released version 2.
DataGridXL is a free (and commercial) editable data table library written in ES6.
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.\n- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.\n- It has its own Virtual DOM implementation to prevent DOM errors.\n- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there. And let us know if you have any suggestions.
What's new in version 2:
* Rapidly search cells via top bar, just like Google Sheets.\n* Freeze any amount of rows and/or columns.\n* Hide and unhide rows and/or columns, just like in Google Sheets.\n* Multi-Range Selection\n* Fullscreen Mode\n* Input bar at the bottom, so it works on touch devices too.
DataGridXL on Github: https://github.com/DataGridXL/DataGridXL2\nDataGridXL on NPM: https://www.npmjs.com/package/@datagridxl/datagridxl2\nDataGridXL Docs: https://datagridxl.com/docs
Please let me know if you have any suggestions or comments!","title":"Show HN: Datagridxl2.js \u2013 No-nonsense fast Excel-like data table library","updated_at":"2024-09-20T10:35:32Z","url":"https://www.datagridxl.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"SchizoDuckie"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I've been, like many of you, researching all the new shiny JS frameworks coming out lately, and found some things very lacking in most, if not all of them:
* Local SQL Data access!\n * Easy CRUD\n * A simple rails-like api for finding relational data\nMost of the projects listed have a way of reading and publishing objects to an API endpoint, but none of them seem to be able to easily operate offline to SQLite in a webview.
My project aims to solve this, in a framework agnostic way, and the whole thing stays under 10kb when compressed.
It's still in a very premature state, has no bootstrap yet and works only on predefined databases for now, but the basics and find API already work. Check out the tests.js for some examples, I'd love to hear some feedback on this.
https://github.com/SchizoDuckie/jsdbobject"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: I'm porting my tiny PHP ORM/ActiveRecord lib to pure Javascript"},"url":{"matchLevel":"none","matchedWords":[],"value":""}},"_tags":["story","author_SchizoDuckie","story_4307248","show_hn"],"author":"SchizoDuckie","children":[4307249],"created_at":"2012-07-29T02:05:00Z","created_at_i":1343527500,"num_comments":1,"objectID":"4307248","points":1,"story_id":4307248,"story_text":"I've been, like many of you, researching all the new shiny JS frameworks coming out lately, and found some things very lacking in most, if not all of them:
* Local SQL Data access!\n * Easy CRUD\n * A simple rails-like api for finding relational data\nMost of the projects listed have a way of reading and publishing objects to an API endpoint, but none of them seem to be able to easily operate offline to SQLite in a webview.
My project aims to solve this, in a framework agnostic way, and the whole thing stays under 10kb when compressed.
It's still in a very premature state, has no bootstrap yet and works only on predefined databases for now, but the basics and find API already work. Check out the tests.js for some examples, I'd love to hear some feedback on this.
https://github.com/SchizoDuckie/jsdbobject","title":"Show HN: I'm porting my tiny PHP ORM/ActiveRecord lib to pure Javascript","updated_at":"2024-09-19T18:42:00Z","url":""},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"irnco"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I have a small bakery called Delis Croissants in Amol, Iran. I beleive we have the most delicious croissants in the world so we need a cool digital menu as well but I couldn't find a digital menu I liked, so I asked claude code to built my own: Rumizi. Rumizi is Persian for tablecloth, the thing on the table.
Since you probably can't visit my bakery, I made demos from publicly available data for top NYC restaurants \u2014 starting with Katz's Delicatessen, which makes the best sandwiches in the world: https://katz-s-delicatessen.rumizi.com/
More NYC spots here: https://rumizi.com/restaurants
each menu has a chat assistant, so you can ask it what to order.\nI'm not affiliated with any of these restaurants; the demos are built from public data. I'm adding more, if you want your favorite restaurant on Rumizi, tell me here, or @irnco on Telegram or @srasoulj on Twitter."},"title":{"matchLevel":"none","matchedWords":[],"value":"What if restaurant menus had AI copilots?"}},"_tags":["story","author_irnco","story_48957135","ask_hn"],"author":"irnco","created_at":"2026-07-18T11:40:33Z","created_at_i":1784374833,"num_comments":0,"objectID":"48957135","points":1,"story_id":48957135,"story_text":"I have a small bakery called Delis Croissants in Amol, Iran. I beleive we have the most delicious croissants in the world so we need a cool digital menu as well but I couldn't find a digital menu I liked, so I asked claude code to built my own: Rumizi. Rumizi is Persian for tablecloth, the thing on the table.
Since you probably can't visit my bakery, I made demos from publicly available data for top NYC restaurants \u2014 starting with Katz's Delicatessen, which makes the best sandwiches in the world: https://katz-s-delicatessen.rumizi.com/
More NYC spots here: https://rumizi.com/restaurants
each menu has a chat assistant, so you can ask it what to order.\nI'm not affiliated with any of these restaurants; the demos are built from public data. I'm adding more, if you want your favorite restaurant on Rumizi, tell me here, or @irnco on Telegram or @srasoulj on Twitter.","title":"What if restaurant menus had AI copilots?","updated_at":"2026-07-18T11:42:15Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"dipampaul17"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I built a CLI that scores robot demonstration episodes using mutual information between states and actions.
The problem: robot learning datasets contain bad demos (jerky movements, hesitation, inconsistent timing). Training on these hurts policy performance. Manual review doesn't scale.
pip install democlean\n democlean analyze lerobot/pusht\nValidation: I correlated MI scores with motion metrics on lerobot/pusht (human teleoperation data). High-MI episodes had 12% lower jerk (p=0.02) and 24% higher state-action correlation (p=0.03). I did not train policies to measure downstream improvement.
Limitations I want to be upfront about:
- MI correlates with episode length (r\u22480.8). Longer episodes score higher.
- This measures motion smoothness, not task success.
- Works best with 50+ episodes from a single task.
- Inspired by DemInf (Hejna et al., RSS 2025) but uses raw KSG estimation instead of their VAE pipeline. Simpler, probably less accurate for high-dimensional observations.
Complements score_lerobot_episodes which catches visual issues (blur, lighting). This catches behavioral issues.
GitHub: https://github.com/dipampaul17/democlean
Happy to answer questions about the approach or validation."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["delos"],"value":"Show HN: Democlean \u2013 Score robot demos by motion quality"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/dipampaul17/democlean"}},"_tags":["story","author_dipampaul17","story_46851191","show_hn"],"author":"dipampaul17","created_at":"2026-02-02T01:10:35Z","created_at_i":1769994635,"num_comments":0,"objectID":"46851191","points":1,"story_id":46851191,"story_text":"I built a CLI that scores robot demonstration episodes using mutual information between states and actions.
The problem: robot learning datasets contain bad demos (jerky movements, hesitation, inconsistent timing). Training on these hurts policy performance. Manual review doesn't scale.
pip install democlean\n democlean analyze lerobot/pusht\nValidation: I correlated MI scores with motion metrics on lerobot/pusht (human teleoperation data). High-MI episodes had 12% lower jerk (p=0.02) and 24% higher state-action correlation (p=0.03). I did not train policies to measure downstream improvement.
Limitations I want to be upfront about:
- MI correlates with episode length (r\u22480.8). Longer episodes score higher.
- This measures motion smoothness, not task success.
- Works best with 50+ episodes from a single task.
- Inspired by DemInf (Hejna et al., RSS 2025) but uses raw KSG estimation instead of their VAE pipeline. Simpler, probably less accurate for high-dimensional observations.
Complements score_lerobot_episodes which catches visual issues (blur, lighting). This catches behavioral issues.
GitHub: https://github.com/dipampaul17/democlean
Happy to answer questions about the approach or validation.","title":"Show HN: Democlean \u2013 Score robot demos by motion quality","updated_at":"2026-03-05T23:28:55Z","url":"https://github.com/dipampaul17/democlean"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"robbiejs"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I'm Robbert, the creator of DataGridXL.js.
DataGridXL is a free (and commercial) editable data table library written in ES6.
DataGridXL has customers from 4 different continents. It is used by Piktochart among many others. It has over 10 million end users, something that I am super proud of!
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.
- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.
- It has its own Virtual DOM implementation to prevent DOM errors.
- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there.
And let us know if you have any suggestions.
I am working on version 3 that will come with even more features, like dropdowns, cell validation etc.
Please let me know if you have any suggestions or comments."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: Datagridxl2.js \u2013 fast Excel-like data table library"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"https://www.datagridxl.com/"}},"_tags":["story","author_robbiejs","story_37836646","show_hn"],"author":"robbiejs","created_at":"2023-10-10T19:54:05Z","created_at_i":1696967645,"num_comments":0,"objectID":"37836646","points":1,"story_id":37836646,"story_text":"I'm Robbert, the creator of DataGridXL.js.
DataGridXL is a free (and commercial) editable data table library written in ES6.
DataGridXL has customers from 4 different continents. It is used by Piktochart among many others. It has over 10 million end users, something that I am super proud of!
My goal is to develop the most performant & user-friendly spreadsheet-like data table out there:
- It has zero dependencies. You don\u2019t need any framework to use DataGridXL.
- It is lightweight (~250kb) and easy to use. It does not even require messing with CSS.
- It has its own Virtual DOM implementation to prevent DOM errors.
- Developer friendly. Supports all modern web browsers
Please take a look at the performance demo (https://www.datagridxl.com/demos/one-million-cells) to see the difference with other data grids out there.
And let us know if you have any suggestions.
I am working on version 3 that will come with even more features, like dropdowns, cell validation etc.
Please let me know if you have any suggestions or comments.","title":"Show HN: Datagridxl2.js \u2013 fast Excel-like data table library","updated_at":"2024-09-20T15:23:03Z","url":"https://www.datagridxl.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"arunmoezhi"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Getting Started with Graph Databases"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"https://academy.datastax.com/demos/getting-started-graph-databases"}},"_tags":["story","author_arunmoezhi","story_11257643"],"author":"arunmoezhi","children":[11257842,11257844,11257892,11257904,11257915,11258026,11258295,11258440,11258453,11258831,11260181],"created_at":"2016-03-10T06:27:46Z","created_at_i":1457591266,"num_comments":32,"objectID":"11257643","points":232,"story_id":11257643,"title":"Getting Started with Graph Databases","updated_at":"2024-09-19T22:57:44Z","url":"https://academy.datastax.com/demos/getting-started-graph-databases"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"DatMule"},"title":{"matchLevel":"none","matchedWords":[],"value":"Rookie Move - links work but not to intended demo, worth lookin tho. my bad."},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"https://www.DataMolecular.com/demos"}},"_tags":["story","author_DatMule","story_45743536"],"author":"DatMule","children":[45743537,45743577],"created_at":"2025-10-29T07:03:54Z","created_at_i":1761721434,"num_comments":1,"objectID":"45743536","points":1,"story_id":45743536,"title":"Rookie Move - links work but not to intended demo, worth lookin tho. my bad.","updated_at":"2026-03-05T22:57:33Z","url":"https://www.DataMolecular.com/demos"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"lumbroso"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"This is an open\u2011source Model Context Protocol (MCP) server that gives any LLM a sense of the passage of time.
Most MCP demos wire LLMs to external data stores. That\u2019s useful, but MCP is also a chance to give models perception \u2014 extra senses beyond the prompt text.
Six functions (`current_datetime`, `time_difference`, `timestamp_context`, etc.) give Claude/GPT real temporal awareness: It can spot pauses, reason about rhythms, and even label a chat\u2019s \u201cthree\u2011act structure\u201d. Runs locally in <60 s (Python) or via a hosted demo.
If time works, what else could we surface? - Location / movement (GPS, speed, \u201cI\u2019m on a train\u201d) - Weather (rainy evening vs clear morning) - Device state (battery low, poor bandwidth) - Ambient modality (user is dictating on mobile vs typing at desk) - Calendar context (meeting starts in 5 min) - Biometric cues (heart\u2011rate spikes while coding)
Curious what other signals people think would unlock better collaboration.
Full back story: https://medium.com/@jeremie.lumbroso/teaching-ai-the-signifi...
Happy to discuss MCP patterns, tool discovery, or future \u201csenses\u201d. Feedback and PRs welcome!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: An MCP server that gives LLMs temporal awareness and time calculation"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/jlumbroso/passage-of-time-mcp"}},"_tags":["story","author_lumbroso","story_44583014","show_hn"],"author":"lumbroso","children":[44583021,44583414,44583582,44583586,44583603,44583695,44583762,44583787,44583863,44583957,44585214,44586413,44587090,44588850],"created_at":"2025-07-16T14:55:31Z","created_at_i":1752677731,"num_comments":55,"objectID":"44583014","points":91,"story_id":44583014,"story_text":"This is an open\u2011source Model Context Protocol (MCP) server that gives any LLM a sense of the passage of time.
Most MCP demos wire LLMs to external data stores. That\u2019s useful, but MCP is also a chance to give models perception \u2014 extra senses beyond the prompt text.
Six functions (`current_datetime`, `time_difference`, `timestamp_context`, etc.) give Claude/GPT real temporal awareness: It can spot pauses, reason about rhythms, and even label a chat\u2019s \u201cthree\u2011act structure\u201d. Runs locally in <60 s (Python) or via a hosted demo.
If time works, what else could we surface? - Location / movement (GPS, speed, \u201cI\u2019m on a train\u201d) - Weather (rainy evening vs clear morning) - Device state (battery low, poor bandwidth) - Ambient modality (user is dictating on mobile vs typing at desk) - Calendar context (meeting starts in 5 min) - Biometric cues (heart\u2011rate spikes while coding)
Curious what other signals people think would unlock better collaboration.
Full back story: https://medium.com/@jeremie.lumbroso/teaching-ai-the-signifi...
Happy to discuss MCP patterns, tool discovery, or future \u201csenses\u201d. Feedback and PRs welcome!","title":"Show HN: An MCP server that gives LLMs temporal awareness and time calculation","updated_at":"2026-09-24T17:24:24Z","url":"https://github.com/jlumbroso/passage-of-time-mcp"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"a6kme"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hi HN, I have been building voice agents for sometime now. I was earlier automating parts of visa processing, and we needed real-time, multilingual voice calling.
I assumed the hard work was just wiring LiveKit/Pipecat + STT/TTS + an LLM. It wasn\u2019t.
Even with solid OSS (Pipecat/LiveKit), we still had to do a lot of plumbing- variable extraction, tracing, testing etc and any workflow changes required constant redeploys.
We eventually realized we\u2019d spent more time building infrastructure than building the actual agents. Everything felt custom. We hit every possible pain with Pipecat and VAPI style systems.
So we built Dograh - a fully open-source voice agent framework that includes all the boring, painful pieces by default.
What\u2019s different:
- Pipecat-based engine, but forked - custom event model, and concurrency fixes
- One-click start template generated by an LLM Agent for a quick get start template for any use case
- Drag-and-drop visual agent builder for quick iteration (the thing we wished existed earlier)
- Variable extraction layer (name/order/date/etc.) baked into the LLM loop
- Built in Telephony integration (Twilio/ Vonage/ Vobiz/ Cloudonix)
- Multilingual support end-to-end
- Select any LLM TTS STT (add their credits, if any)
- AI-to-AI call testing: automatically stress-test an agent before shipping (still a work in progress- so patchy as of now)
- Fully Open Source
It's built and maintained by YC alumni / exit founders who got tired of rebuilding the same plumbing.
Why we open-sourced it:\nWe kept feeling that the space was drifting toward closed SaaS abstractions (VAPI, Retell). Those are good for demos, but once you need data controls, privacy or self/offline deployment, you end up stuck. We wanted a stack where you can see every part, fork it, self-host it, and patch it as needed.
Try it:
- Repo: https://github.com/dograh-hq/dograh
This spins up a basic multilingual agent with everything pre-wired.
Who this is for:
- If you are looking for self hosting a Vapi like platform for Data Privacy etc.
- Anyone trying to build production-grade voice agents without reinventing audio plumbing.
- If you\u2019ve tried to glue STT\u2192LLM\u2192TTS manually, you probably know the exact pain this is built for
Happy to answer technical questions, show the architecture, or hear how we can improve the product."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Dograh \u2013 an OSS Vapi alternative to quickly build and test voice agents"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/dograh-hq/dograh"}},"_tags":["story","author_a6kme","story_46189836","show_hn"],"author":"a6kme","children":[46189900,46189953,46270202,46270222,46270301,46270751],"created_at":"2025-12-08T08:35:24Z","created_at_i":1765182924,"num_comments":10,"objectID":"46189836","points":16,"story_id":46189836,"story_text":"Hi HN, I have been building voice agents for sometime now. I was earlier automating parts of visa processing, and we needed real-time, multilingual voice calling.
I assumed the hard work was just wiring LiveKit/Pipecat + STT/TTS + an LLM. It wasn\u2019t.
Even with solid OSS (Pipecat/LiveKit), we still had to do a lot of plumbing- variable extraction, tracing, testing etc and any workflow changes required constant redeploys.
We eventually realized we\u2019d spent more time building infrastructure than building the actual agents. Everything felt custom. We hit every possible pain with Pipecat and VAPI style systems.
So we built Dograh - a fully open-source voice agent framework that includes all the boring, painful pieces by default.
What\u2019s different:
- Pipecat-based engine, but forked - custom event model, and concurrency fixes
- One-click start template generated by an LLM Agent for a quick get start template for any use case
- Drag-and-drop visual agent builder for quick iteration (the thing we wished existed earlier)
- Variable extraction layer (name/order/date/etc.) baked into the LLM loop
- Built in Telephony integration (Twilio/ Vonage/ Vobiz/ Cloudonix)
- Multilingual support end-to-end
- Select any LLM TTS STT (add their credits, if any)
- AI-to-AI call testing: automatically stress-test an agent before shipping (still a work in progress- so patchy as of now)
- Fully Open Source
It's built and maintained by YC alumni / exit founders who got tired of rebuilding the same plumbing.
Why we open-sourced it:\nWe kept feeling that the space was drifting toward closed SaaS abstractions (VAPI, Retell). Those are good for demos, but once you need data controls, privacy or self/offline deployment, you end up stuck. We wanted a stack where you can see every part, fork it, self-host it, and patch it as needed.
Try it:
- Repo: https://github.com/dograh-hq/dograh
This spins up a basic multilingual agent with everything pre-wired.
Who this is for:
- If you are looking for self hosting a Vapi like platform for Data Privacy etc.
- Anyone trying to build production-grade voice agents without reinventing audio plumbing.
- If you\u2019ve tried to glue STT\u2192LLM\u2192TTS manually, you probably know the exact pain this is built for
Happy to answer technical questions, show the architecture, or hear how we can improve the product.","title":"Show HN: Dograh \u2013 an OSS Vapi alternative to quickly build and test voice agents","updated_at":"2026-05-17T21:36:07Z","url":"https://github.com/dograh-hq/dograh"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"endless-r0ad"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Good morning!
As an avid quake 3 defrag player, this project started out for me as a tool for Quake 3 and Q3 mods that would solve some of the issues that the in-game UI has. It is essentially an external launcher for the game built with Tauri v2.
Where this launcher shines is it lets you toggle your Quake 3 client exe and with 1 click the launcher is now targeted to that mod (servers / demos / levels refresh). Note there is a nice GIF of the app in the readme of the repo
The project uses Tauri (webview) which some people scoff at - and I get it. Although I think most of the webview stigma is due to electron apps bloating everyone's PC. In any case - I tried my damndest on the UI :).
So try it out - this is for all the Q3 old heads!
Benefits over in-game UI below:
Faster server browser, \nAbility to favorite servers, \nAbility to trash servers (this is useful as many servers advertise using fake players which pollutes the in-game browsing experience - trashing these resolves this),\nAbility to add custom servers that are not advertised on the Q3 master servers, \nUnlimited number of demos/levels can be displayed (Q3 client exes have hard limits on the amount of demos / levels that can be displayed in the in-game UI),\nAbility to play demos on a loop (a feature inspired by the old Demo Show Creator tool https://alt.3dcenter.org/dsc/index_english.php), \nAbility to auto-close the client at demo end, \nManaging the EXE (basically closing and re-opening the client on subsequent launches), \nMore data is parsed from demos, levels, and servers and is displayed in a nice fashion, \nYou can read back the console activity in demos (chat history etc) without having to play the demo. I don't think this has been done before in any other tool"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Sarge Launcher \u2013 A Quake 3 Arena Utility"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/endless-r0ad/sarge-launcher"}},"_tags":["story","author_endless-r0ad","story_45518016","show_hn"],"author":"endless-r0ad","children":[45518386],"created_at":"2025-10-08T16:36:41Z","created_at_i":1759941401,"num_comments":2,"objectID":"45518016","points":5,"story_id":45518016,"story_text":"Good morning!
As an avid quake 3 defrag player, this project started out for me as a tool for Quake 3 and Q3 mods that would solve some of the issues that the in-game UI has. It is essentially an external launcher for the game built with Tauri v2.
Where this launcher shines is it lets you toggle your Quake 3 client exe and with 1 click the launcher is now targeted to that mod (servers / demos / levels refresh). Note there is a nice GIF of the app in the readme of the repo
The project uses Tauri (webview) which some people scoff at - and I get it. Although I think most of the webview stigma is due to electron apps bloating everyone's PC. In any case - I tried my damndest on the UI :).
So try it out - this is for all the Q3 old heads!
Benefits over in-game UI below:
Faster server browser, \nAbility to favorite servers, \nAbility to trash servers (this is useful as many servers advertise using fake players which pollutes the in-game browsing experience - trashing these resolves this),\nAbility to add custom servers that are not advertised on the Q3 master servers, \nUnlimited number of demos/levels can be displayed (Q3 client exes have hard limits on the amount of demos / levels that can be displayed in the in-game UI),\nAbility to play demos on a loop (a feature inspired by the old Demo Show Creator tool https://alt.3dcenter.org/dsc/index_english.php), \nAbility to auto-close the client at demo end, \nManaging the EXE (basically closing and re-opening the client on subsequent launches), \nMore data is parsed from demos, levels, and servers and is displayed in a nice fashion, \nYou can read back the console activity in demos (chat history etc) without having to play the demo. I don't think this has been done before in any other tool","title":"Show HN: Sarge Launcher \u2013 A Quake 3 Arena Utility","updated_at":"2026-03-05T22:48:01Z","url":"https://github.com/endless-r0ad/sarge-launcher"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Tanvir3"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"I made this searchable database of viral product demos because I\u2019m constantly looking for ideas to promote my SaaS
I usually do this by bookmarking tons of posts on Instagram/TikTok
But that wasn\u2019t enough since I wanted to filter by categories like video type (faceless, talking head, no speech), industry, product, etc
So I built this curated searchable database to make it easier to find inspiration"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Show HN: A searchable database of successful TikTok product demos"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["delos"],"value":"https://demos.firehooks.io/"}},"_tags":["story","author_Tanvir3","story_45056854","show_hn"],"author":"Tanvir3","children":[45059252],"created_at":"2025-08-28T20:47:30Z","created_at_i":1756414050,"num_comments":1,"objectID":"45056854","points":3,"story_id":45056854,"story_text":"I made this searchable database of viral product demos because I\u2019m constantly looking for ideas to promote my SaaS
I usually do this by bookmarking tons of posts on Instagram/TikTok
But that wasn\u2019t enough since I wanted to filter by categories like video type (faceless, talking head, no speech), industry, product, etc
So I built this curated searchable database to make it easier to find inspiration","title":"Show HN: A searchable database of successful TikTok product demos","updated_at":"2026-04-24T05:01:44Z","url":"https://demos.firehooks.io/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mortensonsam"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Hello all!
This is my first time sharing a project here - I built ripoff, a fake data generator for PostgreSQL that goes from templated yaml files to rows in your database.
In my personal and professional work I found that fake data was usually either generated in the application layer, which is awkward and slow(er than SQL), or way too focused on random generation resulting in a database full of data that\u2019s unsavory for humans.
ripoff is built for cases where you know the shape of your data, like local development, integration testing, and setting up demos. Unlike other fake data generators ripoff isn\u2019t aware of your schema or app, so it feels more like writing templated SQL than using a DSL.
The yaml format is a map of unique identifiers to column values, where column values can be literal strings, references to other rows, or functions that generate random data.
All random data generated by ripoff requires an explicit seed, which is neat because re-running ripoff will always generate the same content. That determinism enables ripoff to perform upserts when re-run on the same database, so you don't have to wipe your DB after editing fake data.
Here\u2019s a complex but real world example of what a ripoff file might look like: https://github.com/mortenson/ripoff/blob/main/testdata/real_...
I just published it so there\u2019s some rough edges, but you should be able to give it a try with any project that uses PostgreSQL! Information on installation and use can be found in the README.
Thanks and let me know what you think!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Show HN: Ripoff \u2013 Generate fake data in PostgreSQL from YAML"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/mortenson/ripoff"}},"_tags":["story","author_mortensonsam","story_40838482","show_hn"],"author":"mortensonsam","created_at":"2024-06-30T17:19:05Z","created_at_i":1719767945,"num_comments":0,"objectID":"40838482","points":2,"story_id":40838482,"story_text":"Hello all!
This is my first time sharing a project here - I built ripoff, a fake data generator for PostgreSQL that goes from templated yaml files to rows in your database.
In my personal and professional work I found that fake data was usually either generated in the application layer, which is awkward and slow(er than SQL), or way too focused on random generation resulting in a database full of data that\u2019s unsavory for humans.
ripoff is built for cases where you know the shape of your data, like local development, integration testing, and setting up demos. Unlike other fake data generators ripoff isn\u2019t aware of your schema or app, so it feels more like writing templated SQL than using a DSL.
The yaml format is a map of unique identifiers to column values, where column values can be literal strings, references to other rows, or functions that generate random data.
All random data generated by ripoff requires an explicit seed, which is neat because re-running ripoff will always generate the same content. That determinism enables ripoff to perform upserts when re-run on the same database, so you don't have to wipe your DB after editing fake data.
Here\u2019s a complex but real world example of what a ripoff file might look like: https://github.com/mortenson/ripoff/blob/main/testdata/real_...
I just published it so there\u2019s some rough edges, but you should be able to give it a try with any project that uses PostgreSQL! Information on installation and use can be found in the README.
Thanks and let me know what you think!","title":"Show HN: Ripoff \u2013 Generate fake data in PostgreSQL from YAML","updated_at":"2024-09-20T17:22:14Z","url":"https://github.com/mortenson/ripoff"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"long_time_gone"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Trying to understand how often folks have to create fake or dummy data for product development, demos, proofs-of-concept, consulting, or other purposes.
What tools do you use?"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["data"],"value":"Ask HN: How often do you create fake data sets?"}},"_tags":["story","author_long_time_gone","story_29094260","ask_hn"],"author":"long_time_gone","children":[29094331,29094454,29096963],"created_at":"2021-11-03T13:26:58Z","created_at_i":1635946018,"num_comments":5,"objectID":"29094260","points":1,"story_id":29094260,"story_text":"Trying to understand how often folks have to create fake or dummy data for product development, demos, proofs-of-concept, consulting, or other purposes.
What tools do you use?","title":"Ask HN: How often do you create fake data sets?","updated_at":"2024-09-20T09:45:56Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"patientplatypus"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["delos","data"],"value":"Say I have a program that is running in webassembly in my browser and I want to take input that the user puts in that program and have it output to other parts of my program. For example, say I want to make a really graphically fancy input box in unity (not that this would ever be practical but just as an example) - can unity take this input that I type in the input box and then output it through webassembly? I could see this as being very cool for custom components, but all the webassembly demos I've seen only take data in to the compiled program (a la flash player or similar).
Thanks!
PS I've asked on SO too, but this is a more open ended question that I think would move we'll with the community here. General webassembly discussion for n00bs also is cool!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Can webassembly push output to JavaScript?"}},"_tags":["story","author_patientplatypus","story_16193482","ask_hn"],"author":"patientplatypus","children":[16194017],"created_at":"2018-01-20T14:37:10Z","created_at_i":1516459030,"num_comments":1,"objectID":"16193482","points":1,"story_id":16193482,"story_text":"Say I have a program that is running in webassembly in my browser and I want to take input that the user puts in that program and have it output to other parts of my program. For example, say I want to make a really graphically fancy input box in unity (not that this would ever be practical but just as an example) - can unity take this input that I type in the input box and then output it through webassembly? I could see this as being very cool for custom components, but all the webassembly demos I've seen only take data in to the compiled program (a la flash player or similar).
Thanks!
PS I've asked on SO too, but this is a more open ended question that I think would move we'll with the community here. General webassembly discussion for n00bs also is cool!","title":"Can webassembly push output to JavaScript?","updated_at":"2024-09-20T01:57:56Z"}],"hitsPerPage":50,"nbHits":316,"nbPages":7,"page":0,"params":"query=Delos+Data&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":12,"processingTimingsMS":{"_request":{"queue":1,"roundTrip":22},"afterFetch":{"format":{"highlighting":2,"total":3},"merge":{"mergeLoop":{"prepareNextHit":2,"total":2},"total":3},"total":3},"fetch":{"query":4,"scanning":3,"total":8},"total":12},"query":"Delos Data","serverTimeMS":16}