-
Notifications
You must be signed in to change notification settings - Fork 0
Implement provider adapter layer (direct API or via LiteLLM) #5
Copy link
Copy link
Closed
Labels
prio:criticalBlocks other work, must do firstBlocks other work, must do firstscope:large3+ days of work3+ days of workspec:providersDESIGN_SPEC Section 9 - Model Provider LayerDESIGN_SPEC Section 9 - Model Provider Layertype:featureNew feature implementationNew feature implementation
Milestone
Description
Context
Based on the LiteLLM evaluation (#4), implement the actual provider adapter layer. This is the concrete implementation of the unified interface designed in #3, connecting the framework to real LLM APIs.
Acceptance Criteria
- Anthropic provider working (Claude models)
- At least one additional provider working (OpenRouter or Ollama)
- Implements the unified interface from Design unified provider interface (base classes, protocols) #3
- Provider configuration loaded from YAML config files
- API keys sourced from environment variables
- Full async support for all provider operations
- Integration tests with mock/recorded API responses
- Error handling for: rate limits (429), authentication errors, timeouts
- Clean separation between provider-specific and generic code
Dependencies
- Depends on Evaluate LiteLLM: integration prototype, limitations, alternatives #4 (LiteLLM evaluation and decision)
Design Spec Reference
Section 9.1, 9.3 — Provider Interface and LiteLLM Integration
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
prio:criticalBlocks other work, must do firstBlocks other work, must do firstscope:large3+ days of work3+ days of workspec:providersDESIGN_SPEC Section 9 - Model Provider LayerDESIGN_SPEC Section 9 - Model Provider Layertype:featureNew feature implementationNew feature implementation