calum/calhta-ai
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
No description
11 commits 1 branch 0 tags 717 KiB
  • Python 77.5%
  • TypeScript 18.4%
  • CSS 2.2%
  • Shell 1.3%
  • JavaScript 0.3%
  • Other 0.3%
Find a file
Exact
Calum Robinson ad330e11f1 Close out Phase 5.1 documentation
2026-07-22 06:13:43 +04:00
backend Make inference results live and observable 2026-07-22 04:14:07 +04:00
docs Close out Phase 5.1 documentation 2026-07-22 06:13:43 +04:00
frontend Make inference results live and observable 2026-07-22 04:14:07 +04:00
inference Make inference results live and observable 2026-07-22 04:14:07 +04:00
packaging Make inference results live and observable 2026-07-22 04:14:07 +04:00
scripts Integrate private inference runtime and GPU scheduling 2026-07-22 00:21:46 +04:00
tests Establish calhta AI project foundation 2026-07-21 20:53:16 +04:00
tools Establish calhta AI project foundation 2026-07-21 20:53:16 +04:00
var Establish calhta AI project foundation 2026-07-21 20:53:16 +04:00
.env.example Publish the operator UI through trusted TLS proxy 2026-07-22 03:01:01 +04:00
.gitignore Deliver the operator web interface 2026-07-22 02:32:35 +04:00
AGENTS.md Establish calhta AI project foundation 2026-07-21 20:53:16 +04:00
alembic.ini Establish backend service foundation 2026-07-21 21:47:48 +04:00
CONTINUATION.md Close out Phase 5.1 documentation 2026-07-22 06:13:43 +04:00
pylock.toml Make inference results live and observable 2026-07-22 04:14:07 +04:00
pyproject.toml Make inference results live and observable 2026-07-22 04:14:07 +04:00
README.md Close out Phase 5.1 documentation 2026-07-22 06:13:43 +04:00
requirements.lock Make inference results live and observable 2026-07-22 04:14:07 +04:00

README.md

calhta AI

calhta AI is a private, self-hosted AI API platform for adding targeted AI capabilities to calhta and LOUD applications. It is not intended to replace ChatGPT or provide a general public chatbot.

Applications integrate with calhta AI through either an OpenAI-compatible API or a native asynchronous jobs API. Inference runtimes such as Ollama are private implementation details and may be replaced without changing client integrations.

Initial deployment

  • Host: Debian 13 server at 10.76.50.162
  • Public URL: https://ai.calhta.com
  • Reverse proxy: Nginx Proxy Manager at 10.76.50.254
  • GPU: NVIDIA RTX A400, 4 GB VRAM
  • Repository path: /opt/calhta-ai

Repository layout

Path Purpose
backend/ FastAPI application, migrations and backend tests
frontend/ React/TypeScript web application
inference/ Runtime-neutral inference adapters
packaging/systemd/ Debian systemd units
scripts/ Installation, maintenance and operational scripts
docs/ Product and engineering documentation
tests/ Cross-component and integration tests
tools/ Developer tools
var/ Local runtime data; ignored by Git

Documentation

Start with Product Requirements, Architecture, and Roadmap. Contributors and coding agents must read AGENTS.md before making changes.

Project status

Version 0.1.1 is deployed at https://ai.calhta.com from application commit 22b1291d81927952cb8db302cb28d1361f9fa375, with live database revision 20260722_0005. The React operator UI, native identity and job APIs, durable result handling, private Ollama adapter, model scheduler and independent GPU telemetry are active behind NPM/TLS.

Phase 5.1 corrected result persistence, structured-output validation, finish/truncation metadata and GPU rolling-peak visibility. Its dashboard SSE implementation is not accepted as complete: the owner still sees a visible refresh/glitch every few seconds. Phase 5.2 must make monitoring and job state update in place without flashes, panel replacement or layout movement before Phase 6.