P3-06: Only OpenAI-compatible chat/completions format supported — no Anthropic/Gemini/local #24

New issue

Closed

opened 2026-06-16 13:57:02 +00:00 by Artur · 0 comments

Artur commented

2026-06-16 13:57:02 +00:00

Owner

Severity: P3 (Low)
File: decider/llm.py

Problem

call_llm() hardcodes the OpenAI chat completions API format:

resp = httpx.post(
    config["api_url"],
    headers={"Authorization": f"Bearer {api_key}", ...},
    json={"model": config["model"], "messages": messages, ...},
)

This means the decider cannot use:

Anthropic Claude (different API format)
Gemini (different API format)
Local llama.cpp (different API format)
Any non-OpenAI-compatible provider

The config claims multi-provider support, but it only supports OpenAI-clones.

Fix

Add provider-specific request formatters, or use a universal adapter like LiteLLM or the Hermes LLM gateway that normalizes API formats.

**Severity**: P3 (Low) **File**: `decider/llm.py` ## Problem `call_llm()` hardcodes the OpenAI chat completions API format: ```python resp = httpx.post( config["api_url"], headers={"Authorization": f"Bearer {api_key}", ...}, json={"model": config["model"], "messages": messages, ...}, ) ``` This means the decider cannot use: - Anthropic Claude (different API format) - Gemini (different API format) - Local llama.cpp (different API format) - Any non-OpenAI-compatible provider The config claims multi-provider support, but it only supports OpenAI-clones. ## Fix Add provider-specific request formatters, or use a universal adapter like LiteLLM or the Hermes LLM gateway that normalizes API formats.

Artur added the

enhancement

P3-Low

labels

2026-06-16 13:57:02 +00:00

Artur referenced this issue from a commit

2026-06-16 14:01:52 +00:00

feat: P3-06 — multi-provider LLM support (OpenAI/Anthropic/Gemini)

Artur closed this issue

2026-06-16 14:01:52 +00:00