name: "Feature: Multi-model LLM support"
about: Allow switching between different LLM backends (Mistral, LLaMA3, GPT, etc.)
title: "[FEATURE] Multi-model LLM support"
labels: enhancement, AI/NLP
Summary
The system currently hardcodes "model": "mistral" in src/llm.py. There is no way to configure or switch the LLM model at runtime or via environment variables.
Problem
- Users on different hardware may not be able to run Mistral locally.
- There is no support for cloud-based models (OpenAI, Anthropic) or other Ollama-hosted models (LLaMA3, Phi3, Gemma).
- Changing the model requires editing source code directly.
Proposed Solution
- Add a
model parameter to the LLM class constructor.
- Read the model name from an environment variable
LLM_MODEL with mistral as default.
- Expose the model selection via the API request schema (
FormFill).
- Document supported models in README.
Acceptance Criteria
Additional Context
Relevant file: src/llm.py, api/schemas/forms.py, src/file_manipulator.py
name: "Feature: Multi-model LLM support"
about: Allow switching between different LLM backends (Mistral, LLaMA3, GPT, etc.)
title: "[FEATURE] Multi-model LLM support"
labels: enhancement, AI/NLP
Summary
The system currently hardcodes
"model": "mistral"insrc/llm.py. There is no way to configure or switch the LLM model at runtime or via environment variables.Problem
Proposed Solution
modelparameter to theLLMclass constructor.LLM_MODELwithmistralas default.FormFill).Acceptance Criteria
LLMclass accepts amodelparameterLLM_MODELenv var, falling back tomistralmodelfieldAdditional Context
Relevant file:
src/llm.py,api/schemas/forms.py,src/file_manipulator.py