WHOOSH-LLM-002: Implement LLM Integration for Team Composition Engine #3

New Issue

tony · 2025-09-21T06:23:42Z

tony commented

2025-09-21 06:23:42 +00:00

Problem

Core LLM integration functions are stubbed out with "not implemented" errors, causing WHOOSH to fall back to basic heuristics only. This blocks the full AI-powered team composition functionality.

Location

File: internal/composer/service.go
Lines: 143 (classifyTaskWithLLM), 421 (analyzeSkillRequirementsWithLLM)

Missing Implementations

1. classifyTaskWithLLM() - Line 143

Current: Returns "not implemented" error
Required: Implement LLM-powered task classification using Ollama API

2. analyzeSkillRequirementsWithLLM() - Line 421

Current: Returns "not implemented" error
Required: Implement LLM-powered skill requirement analysis

Technical Requirements

LLM Integration Architecture

Ollama API Integration:
- Connect to local Ollama instances (192.168.1.{27,72,113,106}:11434)
- Model selection based on task complexity
- Proper error handling and fallback to heuristics
Task Classification:
- Analyze project requirements and classify task types
- Determine complexity levels and required expertise
- Map to predefined role categories
Skill Requirement Analysis:
- Parse project documentation and codebase
- Identify required technical skills and domain knowledge
- Generate skill requirement profiles for team formation

Feature Flag Support

Implementation should respect existing feature flags:

WHOOSH_FEATURE_LLM_CLASSIFICATION - AI vs rule-based classification
WHOOSH_FEATURE_LLM_SKILL_ANALYSIS - AI skill requirement analysis

Priority

High - Core functionality for AI-powered team composition

Requirement Traceability

@goal: WHOOSH-LLM-002, WSH-AUTO - LLM-powered team composition

Dependencies

Ollama API endpoints available on cluster
Team Composer Spec requirements (lines 51-229)
Feature flag configuration system

Implementation Notes

Example Integration Pattern

func (s *Service) classifyTaskWithLLM(ctx context.Context, task Task) (*Classification, error) {
    if !s.config.Features.LLMClassification {
        return s.classifyTaskWithHeuristics(ctx, task)
    }
    
    // Ollama API call
    client := ollama.NewClient(s.config.OllamaEndpoint)
    result, err := client.Generate(ctx, &ollama.GenerateRequest{
        Model: "llama3.1:8b",
        Prompt: s.buildClassificationPrompt(task),
    })
    
    if err != nil {
        // Fallback to heuristics on error
        log.Warn("LLM classification failed, falling back to heuristics", "error", err)
        return s.classifyTaskWithHeuristics(ctx, task)
    }
    
    return s.parseClassificationResponse(result.Response)
}

Acceptance Criteria

classifyTaskWithLLM() connects to Ollama and returns valid classifications
analyzeSkillRequirementsWithLLM() analyzes requirements and returns skill profiles
Feature flags properly control LLM vs heuristic execution
Graceful fallback to heuristics on LLM API failures
Integration tests verify Ollama connectivity
Unit tests cover prompt generation and response parsing
Performance acceptable (< 5s response time for typical requests)

## Problem Core LLM integration functions are stubbed out with "not implemented" errors, causing WHOOSH to fall back to basic heuristics only. This blocks the full AI-powered team composition functionality. ## Location **File:** `internal/composer/service.go` **Lines:** 143 (classifyTaskWithLLM), 421 (analyzeSkillRequirementsWithLLM) ## Missing Implementations ### 1. classifyTaskWithLLM() - Line 143 **Current:** Returns "not implemented" error **Required:** Implement LLM-powered task classification using Ollama API ### 2. analyzeSkillRequirementsWithLLM() - Line 421 **Current:** Returns "not implemented" error **Required:** Implement LLM-powered skill requirement analysis ## Technical Requirements ### LLM Integration Architecture 1. **Ollama API Integration:** - Connect to local Ollama instances (192.168.1.{27,72,113,106}:11434) - Model selection based on task complexity - Proper error handling and fallback to heuristics 2. **Task Classification:** - Analyze project requirements and classify task types - Determine complexity levels and required expertise - Map to predefined role categories 3. **Skill Requirement Analysis:** - Parse project documentation and codebase - Identify required technical skills and domain knowledge - Generate skill requirement profiles for team formation ## Feature Flag Support Implementation should respect existing feature flags: - `WHOOSH_FEATURE_LLM_CLASSIFICATION` - AI vs rule-based classification - `WHOOSH_FEATURE_LLM_SKILL_ANALYSIS` - AI skill requirement analysis ## Priority **High** - Core functionality for AI-powered team composition ## Requirement Traceability @goal: WHOOSH-LLM-002, WSH-AUTO - LLM-powered team composition ## Dependencies - Ollama API endpoints available on cluster - Team Composer Spec requirements (lines 51-229) - Feature flag configuration system ## Implementation Notes ### Example Integration Pattern ```go func (s *Service) classifyTaskWithLLM(ctx context.Context, task Task) (*Classification, error) { if !s.config.Features.LLMClassification { return s.classifyTaskWithHeuristics(ctx, task) } // Ollama API call client := ollama.NewClient(s.config.OllamaEndpoint) result, err := client.Generate(ctx, &ollama.GenerateRequest{ Model: "llama3.1:8b", Prompt: s.buildClassificationPrompt(task), }) if err != nil { // Fallback to heuristics on error log.Warn("LLM classification failed, falling back to heuristics", "error", err) return s.classifyTaskWithHeuristics(ctx, task) } return s.parseClassificationResponse(result.Response) } ``` ## Acceptance Criteria - [ ] classifyTaskWithLLM() connects to Ollama and returns valid classifications - [ ] analyzeSkillRequirementsWithLLM() analyzes requirements and returns skill profiles - [ ] Feature flags properly control LLM vs heuristic execution - [ ] Graceful fallback to heuristics on LLM API failures - [ ] Integration tests verify Ollama connectivity - [ ] Unit tests cover prompt generation and response parsing - [ ] Performance acceptable (< 5s response time for typical requests)

tony added the

bzzz-task

label 2025-09-21 06:29:58 +00:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: tony/WHOOSH#3