Bug: LLM_PROVIDER=ollama undocumented — app crashes with "Unsupported LLM provider: ollama"

[aminrj]

Summary

The README and .env.example both document LLM_PROVIDER=ollama as a valid
option, but setting it causes an immediate crash at startup. The Ollama provider
is referenced in config but never implemented in the LLM client layer.

Steps to reproduce

1. Clone the repo and copy .env.example to .env
2. Set LLM_PROVIDER=ollama and OLLAMA_BASE_URL=http://localhost:11434
3. Run docker compose up

Error

ValueError: Unsupported LLM provider: ollama
File "/app/finbot/core/llm/client.py", line 35, in _get_client
raise ValueError(f"Unsupported LLM provider: {self.provider}")

Root cause

Two separate issues compound each other:

1. finbot/core/llm/client.py has no ollama branch — only openai is handled
2. Even when worked around via LLM_PROVIDER=openai + OPENAI_BASE_URL,
   finbot/agents/chat.py and finbot/core/llm/openai_client.py use the
   OpenAI Responses API (client.responses.create()) which Ollama does not
   implement. Ollama only supports Chat Completions (/v1/chat/completions).

Impact

Anyone wanting to run FinBot locally without an OpenAI API key cannot use the
platform at all, despite the README advertising local Ollama support.

Proposed fix

• Route LLM_PROVIDER=ollama through the OpenAI-compatible client with
  base_url set to OLLAMA_BASE_URL/v1
• Rewrite the two responses.create() call sites to use
  chat.completions.create() with streaming delta handling
• Remap Responses API message formats (function_call, function_call_output,
  developer role) to Chat Completions equivalents
• Document the working .env setup for Ollama in README

I have a working patch and can submit a PR if this approach is acceptable.

Environment

• Ollama 0.x, WSL2 Ubuntu 24.04
• Tested with qwen3.5:9b as the target model
• All challenges functional after patching