feat(context): Phase 5 - Model Adapters (Claude, OpenAI) #83

New Issue

cardosofelipe · 2026-01-04T00:52:20Z

cardosofelipe commented

2026-01-04 00:52:20 +00:00

Overview

Implement model-specific context formatting adapters for optimal LLM interaction.

Parent Issue

#61: Context Management Engine

Implementation Tasks

1. Create `adapters/base.py`

Create ModelAdapter abstract base class
Define abstract format() method
Implement _group_by_type() helper

class ModelAdapter(ABC):
    @abstractmethod
    def format(self, contexts: list[BaseContext]) -> str:
        """Format contexts for the specific model."""
        ...

    def _group_by_type(
        self,
        contexts: list[BaseContext]
    ) -> dict[ContextType, list[BaseContext]]:
        """Group contexts by their type."""
        ...

2. Create `adapters/claude.py`

Create ClaudeAdapter class
Use XML tags for structure (<documents>, <document>, <task>)
Format system context appropriately
Order: System → Knowledge → Task → Conversation

class ClaudeAdapter(ModelAdapter):
    """Claude-specific formatting with XML tags."""

    def format(self, contexts: list[BaseContext]) -> str:
        parts = []
        by_type = self._group_by_type(contexts)

        if ContextType.KNOWLEDGE in by_type:
            parts.append("<documents>")
            for ctx in by_type[ContextType.KNOWLEDGE]:
                parts.append(f'<document source="{ctx.source}">{ctx.content}</document>')
            parts.append("</documents>")
        ...

3. Create `adapters/openai.py`

Create OpenAIAdapter class
Use markdown formatting
Use headers for structure (##, ###)

class OpenAIAdapter(ModelAdapter):
    """OpenAI-specific formatting with markdown."""

    def format(self, contexts: list[BaseContext]) -> str:
        parts = []
        by_type = self._group_by_type(contexts)

        if ContextType.KNOWLEDGE in by_type:
            parts.append("## Reference Documents\n")
            for ctx in by_type[ContextType.KNOWLEDGE]:
                parts.append(f"### Source: {ctx.source}\n{ctx.content}\n")
        ...

4. Create `adapters/default.py`

Create DefaultAdapter for unknown models
Use simple text formatting
Add source attribution

5. Update `adapters/init.py`

Export all adapters
Create get_adapter() factory function

Files to Create

backend/app/services/context/adapters/
├── __init__.py
├── base.py
├── claude.py
├── openai.py
└── default.py

Acceptance Criteria

Claude adapter uses XML tags correctly
OpenAI adapter uses markdown correctly
Default adapter works for unknown models
Context ordering is consistent per model
Unit tests for each adapter
Output is well-formatted and readable

Dependencies

#79 (Phase 1 - Foundation)

Labels

phase-2, context, backend

## Overview Implement model-specific context formatting adapters for optimal LLM interaction. ## Parent Issue - #61: Context Management Engine --- ## Implementation Tasks ### 1. Create `adapters/base.py` - [ ] Create `ModelAdapter` abstract base class - [ ] Define abstract `format()` method - [ ] Implement `_group_by_type()` helper ```python class ModelAdapter(ABC): @abstractmethod def format(self, contexts: list[BaseContext]) -> str: """Format contexts for the specific model.""" ... def _group_by_type( self, contexts: list[BaseContext] ) -> dict[ContextType, list[BaseContext]]: """Group contexts by their type.""" ... ``` ### 2. Create `adapters/claude.py` - [ ] Create `ClaudeAdapter` class - [ ] Use XML tags for structure (`<documents>`, `<document>`, `<task>`) - [ ] Format system context appropriately - [ ] Order: System → Knowledge → Task → Conversation ```python class ClaudeAdapter(ModelAdapter): """Claude-specific formatting with XML tags.""" def format(self, contexts: list[BaseContext]) -> str: parts = [] by_type = self._group_by_type(contexts) if ContextType.KNOWLEDGE in by_type: parts.append("<documents>") for ctx in by_type[ContextType.KNOWLEDGE]: parts.append(f'<document source="{ctx.source}">{ctx.content}</document>') parts.append("</documents>") ... ``` ### 3. Create `adapters/openai.py` - [ ] Create `OpenAIAdapter` class - [ ] Use markdown formatting - [ ] Use headers for structure (`##`, `###`) ```python class OpenAIAdapter(ModelAdapter): """OpenAI-specific formatting with markdown.""" def format(self, contexts: list[BaseContext]) -> str: parts = [] by_type = self._group_by_type(contexts) if ContextType.KNOWLEDGE in by_type: parts.append("## Reference Documents\n") for ctx in by_type[ContextType.KNOWLEDGE]: parts.append(f"### Source: {ctx.source}\n{ctx.content}\n") ... ``` ### 4. Create `adapters/default.py` - [ ] Create `DefaultAdapter` for unknown models - [ ] Use simple text formatting - [ ] Add source attribution ### 5. Update `adapters/__init__.py` - [ ] Export all adapters - [ ] Create `get_adapter()` factory function --- ## Files to Create ``` backend/app/services/context/adapters/ ├── __init__.py ├── base.py ├── claude.py ├── openai.py └── default.py ``` --- ## Acceptance Criteria - [ ] Claude adapter uses XML tags correctly - [ ] OpenAI adapter uses markdown correctly - [ ] Default adapter works for unknown models - [ ] Context ordering is consistent per model - [ ] Unit tests for each adapter - [ ] Output is well-formatted and readable --- ## Dependencies - #79 (Phase 1 - Foundation) ## Labels `phase-2`, `context`, `backend`

cardosofelipe referenced this issue

2026-01-04 00:52:21 +00:00

feat(context): Phase 7 - Main Engine & Integration #85

cardosofelipe referenced this issue

2026-01-04 00:54:11 +00:00

feat(mcp): Context Management Engine #61

cardosofelipe referenced this issue from a commit

2026-01-04 14:23:32 +00:00

feat(context): implement model adapters for Claude and OpenAI (#83)

cardosofelipe closed this issue

2026-01-04 18:53:44 +00:00

Sign in to join this conversation.