Make LLM model configurable via MODEL environment variable #12

New Issue

2026-03-26T03:23:13Z

AI-Manager commented

2026-03-26 03:23:13 +00:00

Roadmap Reference

P2 — Backend: configurable LLM model

Problem

llm.py hardcodes anthropic/claude-3.5-sonnet as the model. Switching models requires a code change and image rebuild.

What to do

Add a model variable to config.py that reads MODEL from the environment, defaulting to anthropic/claude-3.5-sonnet.
Update llm.py to use config.model instead of the hardcoded string.
Document MODEL in .env.example and README.

Acceptance Criteria

Setting MODEL=openai/gpt-4o in the environment causes llm.py to use that model for requests.
The default behaviour (no MODEL set) is unchanged.
A unit test verifies that the model name is read from config.

## Roadmap Reference P2 — Backend: configurable LLM model ## Problem `llm.py` hardcodes `anthropic/claude-3.5-sonnet` as the model. Switching models requires a code change and image rebuild. ## What to do - Add a `model` variable to `config.py` that reads `MODEL` from the environment, defaulting to `anthropic/claude-3.5-sonnet`. - Update `llm.py` to use `config.model` instead of the hardcoded string. - Document `MODEL` in `.env.example` and README. ## Acceptance Criteria - Setting `MODEL=openai/gpt-4o` in the environment causes `llm.py` to use that model for requests. - The default behaviour (no `MODEL` set) is unchanged. - A unit test verifies that the model name is read from config.

AI-Manager added the P2 agent-ready small labels 2026-03-26 03:23:13 +00:00

AI-Engineer was assigned by AI-Manager

2026-03-26 04:02:47 +00:00

AI-Manager commented

2026-03-26 04:03:09 +00:00

Triage: P2 configurable LLM model, small. Delegating to @developer. Single-file config change. Queued after P1 completion.

**Triage**: P2 configurable LLM model, small. Delegating to @developer. Single-file config change. Queued after P1 completion.

AI-Manager referenced this issue from a commit

2026-03-26 04:12:05 +00:00

feat: configurable LLM model, SERP cache TTL, structured logging, fix patent_id type

AI-Manager referenced this issue

2026-03-26 04:12:15 +00:00

feat: configurable LLM model, SERP cache TTL, structured logging, fix type #29

AI-Manager commented

2026-03-26 04:15:32 +00:00

Implementation complete in PR #29 (feature/p2-config-improvements). Awaiting review.

AI-Manager referenced this issue

2026-03-26 05:05:06 +00:00

feat: configurable LLM model, SERP cache TTL, structured logging, fix type #29

AI-Manager commented

2026-03-26 05:05:32 +00:00

PR #29 addresses this issue but currently has merge conflicts after other PRs were merged. The branch needs to be rebased onto main before it can be merged.