Resolves WHOOSH-LLM-002: Replace stubbed LLM functions with full Ollama API integration ## New Features - Full Ollama API integration with automatic endpoint discovery - LLM-powered task classification using configurable models - LLM-powered skill requirement analysis - Graceful fallback to heuristics on LLM failures - Feature flag support for LLM vs heuristic execution - Performance optimization with smaller, faster models (llama3.2:latest) ## Implementation Details - Created OllamaClient with connection pooling and timeout handling - Structured prompt engineering for consistent JSON responses - Robust error handling with automatic failover to heuristics - Comprehensive integration tests validating functionality - Support for multiple Ollama endpoints with health checking ## Performance & Reliability - Timeout configuration prevents hanging requests - Fallback mechanism ensures system reliability - Uses 3.2B parameter model for balance of speed vs accuracy - Graceful degradation when LLM services unavailable ## Files Added - internal/composer/ollama.go: Core Ollama API integration - internal/composer/llm_test.go: Comprehensive integration tests ## Files Modified - internal/composer/service.go: Implemented LLM functions - internal/composer/models.go: Updated config for performance 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
342 lines
12 KiB
Go
342 lines
12 KiB
Go
package composer
|
|
|
|
import (
|
|
"bytes"
|
|
"context"
|
|
"encoding/json"
|
|
"fmt"
|
|
"net/http"
|
|
"strings"
|
|
"time"
|
|
|
|
"github.com/rs/zerolog/log"
|
|
)
|
|
|
|
// OllamaClient provides LLM integration with Ollama instances
|
|
type OllamaClient struct {
|
|
baseURL string
|
|
httpClient *http.Client
|
|
model string
|
|
}
|
|
|
|
// OllamaRequest represents a request to the Ollama API
|
|
type OllamaRequest struct {
|
|
Model string `json:"model"`
|
|
Prompt string `json:"prompt"`
|
|
Stream bool `json:"stream"`
|
|
}
|
|
|
|
// OllamaResponse represents a response from the Ollama API
|
|
type OllamaResponse struct {
|
|
Model string `json:"model"`
|
|
CreatedAt time.Time `json:"created_at"`
|
|
Response string `json:"response"`
|
|
Done bool `json:"done"`
|
|
Context []int `json:"context,omitempty"`
|
|
TotalDuration int64 `json:"total_duration,omitempty"`
|
|
LoadDuration int64 `json:"load_duration,omitempty"`
|
|
PromptEvalCount int `json:"prompt_eval_count,omitempty"`
|
|
PromptEvalDuration int64 `json:"prompt_eval_duration,omitempty"`
|
|
EvalCount int `json:"eval_count,omitempty"`
|
|
EvalDuration int64 `json:"eval_duration,omitempty"`
|
|
}
|
|
|
|
// NewOllamaClient creates a new Ollama client with fallback endpoints
|
|
func NewOllamaClient(model string) *OllamaClient {
|
|
// Default Ollama endpoints from cluster
|
|
endpoints := []string{
|
|
"http://192.168.1.27:11434",
|
|
"http://192.168.1.72:11434",
|
|
"http://192.168.1.113:11434",
|
|
"http://192.168.1.106:11434",
|
|
}
|
|
|
|
// Try to find a working endpoint
|
|
baseURL := endpoints[0] // Default to first endpoint
|
|
httpClient := &http.Client{
|
|
Timeout: 30 * time.Second,
|
|
}
|
|
|
|
// Quick health check to find working endpoint
|
|
for _, endpoint := range endpoints {
|
|
resp, err := httpClient.Get(endpoint + "/api/tags")
|
|
if err == nil && resp.StatusCode == 200 {
|
|
baseURL = endpoint
|
|
resp.Body.Close()
|
|
break
|
|
}
|
|
if resp != nil {
|
|
resp.Body.Close()
|
|
}
|
|
}
|
|
|
|
log.Info().
|
|
Str("base_url", baseURL).
|
|
Str("model", model).
|
|
Msg("Initialized Ollama client")
|
|
|
|
return &OllamaClient{
|
|
baseURL: baseURL,
|
|
httpClient: httpClient,
|
|
model: model,
|
|
}
|
|
}
|
|
|
|
// Generate sends a prompt to Ollama and returns the response
|
|
func (c *OllamaClient) Generate(ctx context.Context, prompt string) (string, error) {
|
|
reqBody := OllamaRequest{
|
|
Model: c.model,
|
|
Prompt: prompt,
|
|
Stream: false,
|
|
}
|
|
|
|
jsonData, err := json.Marshal(reqBody)
|
|
if err != nil {
|
|
return "", fmt.Errorf("failed to marshal request: %w", err)
|
|
}
|
|
|
|
req, err := http.NewRequestWithContext(ctx, "POST", c.baseURL+"/api/generate", bytes.NewBuffer(jsonData))
|
|
if err != nil {
|
|
return "", fmt.Errorf("failed to create request: %w", err)
|
|
}
|
|
|
|
req.Header.Set("Content-Type", "application/json")
|
|
|
|
startTime := time.Now()
|
|
resp, err := c.httpClient.Do(req)
|
|
if err != nil {
|
|
return "", fmt.Errorf("failed to send request: %w", err)
|
|
}
|
|
defer resp.Body.Close()
|
|
|
|
if resp.StatusCode != http.StatusOK {
|
|
return "", fmt.Errorf("ollama API returned status %d", resp.StatusCode)
|
|
}
|
|
|
|
var ollamaResp OllamaResponse
|
|
if err := json.NewDecoder(resp.Body).Decode(&ollamaResp); err != nil {
|
|
return "", fmt.Errorf("failed to decode response: %w", err)
|
|
}
|
|
|
|
duration := time.Since(startTime)
|
|
log.Debug().
|
|
Str("model", c.model).
|
|
Dur("duration", duration).
|
|
Int("eval_count", ollamaResp.EvalCount).
|
|
Msg("Ollama generation completed")
|
|
|
|
return ollamaResp.Response, nil
|
|
}
|
|
|
|
// TaskClassificationPrompt builds a prompt for task classification
|
|
func (c *OllamaClient) BuildTaskClassificationPrompt(input *TaskAnalysisInput) string {
|
|
var prompt strings.Builder
|
|
|
|
prompt.WriteString("You are an expert software project manager analyzing development tasks. ")
|
|
prompt.WriteString("Classify the following task and provide analysis in JSON format.\n\n")
|
|
|
|
prompt.WriteString("Task Details:\n")
|
|
prompt.WriteString(fmt.Sprintf("Title: %s\n", input.Title))
|
|
prompt.WriteString(fmt.Sprintf("Description: %s\n", input.Description))
|
|
|
|
if len(input.Requirements) > 0 {
|
|
prompt.WriteString(fmt.Sprintf("Requirements: %s\n", strings.Join(input.Requirements, ", ")))
|
|
}
|
|
|
|
if len(input.TechStack) > 0 {
|
|
prompt.WriteString(fmt.Sprintf("Tech Stack: %s\n", strings.Join(input.TechStack, ", ")))
|
|
}
|
|
|
|
prompt.WriteString(fmt.Sprintf("Priority: %s\n\n", input.Priority))
|
|
|
|
prompt.WriteString("Analyze this task and respond with valid JSON in this EXACT format:\n")
|
|
prompt.WriteString("{\n")
|
|
prompt.WriteString(" \"task_type\": \"feature_development\",\n")
|
|
prompt.WriteString(" \"complexity_score\": 0.7,\n")
|
|
prompt.WriteString(" \"primary_domains\": [\"backend\", \"api\"],\n")
|
|
prompt.WriteString(" \"secondary_domains\": [\"testing\"],\n")
|
|
prompt.WriteString(" \"estimated_duration_hours\": 8,\n")
|
|
prompt.WriteString(" \"risk_level\": \"medium\",\n")
|
|
prompt.WriteString(" \"required_experience\": \"intermediate\"\n")
|
|
prompt.WriteString("}\n\n")
|
|
|
|
prompt.WriteString("Task types: feature_development, bug_fix, refactoring, security, integration, optimization, maintenance\n")
|
|
prompt.WriteString("Complexity: number between 0.1-1.0\n")
|
|
prompt.WriteString("Duration: integer between 1-40\n")
|
|
prompt.WriteString("Risk: minimal, low, medium, high\n")
|
|
prompt.WriteString("Experience: junior, intermediate, senior\n\n")
|
|
|
|
prompt.WriteString("Respond ONLY with valid JSON, no markdown, no other text.")
|
|
|
|
return prompt.String()
|
|
}
|
|
|
|
// SkillAnalysisPrompt builds a prompt for skill requirement analysis
|
|
func (c *OllamaClient) BuildSkillAnalysisPrompt(input *TaskAnalysisInput, classification *TaskClassification) string {
|
|
var prompt strings.Builder
|
|
|
|
prompt.WriteString("You are an expert technical recruiter analyzing skill requirements for software development tasks. ")
|
|
prompt.WriteString("Analyze the task and provide detailed skill requirements in JSON format.\n\n")
|
|
|
|
prompt.WriteString("Task Details:\n")
|
|
prompt.WriteString(fmt.Sprintf("Title: %s\n", input.Title))
|
|
prompt.WriteString(fmt.Sprintf("Description: %s\n", input.Description))
|
|
prompt.WriteString(fmt.Sprintf("Task Type: %s\n", classification.TaskType))
|
|
prompt.WriteString(fmt.Sprintf("Complexity: %.1f\n", classification.ComplexityScore))
|
|
prompt.WriteString(fmt.Sprintf("Primary Domains: %s\n", strings.Join(classification.PrimaryDomains, ", ")))
|
|
|
|
if len(input.Requirements) > 0 {
|
|
prompt.WriteString(fmt.Sprintf("Requirements: %s\n", strings.Join(input.Requirements, ", ")))
|
|
}
|
|
|
|
if len(input.TechStack) > 0 {
|
|
prompt.WriteString(fmt.Sprintf("Tech Stack: %s\n", strings.Join(input.TechStack, ", ")))
|
|
}
|
|
|
|
prompt.WriteString("\nAnalyze and respond with JSON containing:\n")
|
|
prompt.WriteString("{\n")
|
|
prompt.WriteString(" \"critical_skills\": [\n")
|
|
prompt.WriteString(" {\n")
|
|
prompt.WriteString(" \"domain\": \"skill domain name\",\n")
|
|
prompt.WriteString(" \"min_proficiency\": 0.1-1.0,\n")
|
|
prompt.WriteString(" \"weight\": 0.1-1.0,\n")
|
|
prompt.WriteString(" \"critical\": true\n")
|
|
prompt.WriteString(" }\n")
|
|
prompt.WriteString(" ],\n")
|
|
prompt.WriteString(" \"desirable_skills\": [\n")
|
|
prompt.WriteString(" {\n")
|
|
prompt.WriteString(" \"domain\": \"skill domain name\",\n")
|
|
prompt.WriteString(" \"min_proficiency\": 0.1-1.0,\n")
|
|
prompt.WriteString(" \"weight\": 0.1-1.0,\n")
|
|
prompt.WriteString(" \"critical\": false\n")
|
|
prompt.WriteString(" }\n")
|
|
prompt.WriteString(" ],\n")
|
|
prompt.WriteString(" \"total_skill_count\": 0\n")
|
|
prompt.WriteString("}\n\n")
|
|
|
|
prompt.WriteString("Focus on:\n")
|
|
prompt.WriteString("- Programming languages and frameworks\n")
|
|
prompt.WriteString("- Domain expertise (backend, frontend, devops, etc.)\n")
|
|
prompt.WriteString("- Tools and technologies\n")
|
|
prompt.WriteString("- Soft skills relevant to the task type\n\n")
|
|
|
|
prompt.WriteString("Respond ONLY with valid JSON, no other text.")
|
|
|
|
return prompt.String()
|
|
}
|
|
|
|
// ParseTaskClassificationResponse parses LLM response into TaskClassification
|
|
func (c *OllamaClient) ParseTaskClassificationResponse(response string) (*TaskClassification, error) {
|
|
// Clean the response - remove markdown code blocks if present
|
|
response = strings.TrimSpace(response)
|
|
response = strings.TrimPrefix(response, "```json")
|
|
response = strings.TrimPrefix(response, "```")
|
|
response = strings.TrimSuffix(response, "```")
|
|
response = strings.TrimSpace(response)
|
|
|
|
var result struct {
|
|
TaskType string `json:"task_type"`
|
|
ComplexityScore float64 `json:"complexity_score"`
|
|
PrimaryDomains []string `json:"primary_domains"`
|
|
SecondaryDomains []string `json:"secondary_domains"`
|
|
EstimatedDuration int `json:"estimated_duration_hours"`
|
|
RiskLevel string `json:"risk_level"`
|
|
RequiredExperience string `json:"required_experience"`
|
|
}
|
|
|
|
if err := json.Unmarshal([]byte(response), &result); err != nil {
|
|
return nil, fmt.Errorf("failed to parse classification response: %w", err)
|
|
}
|
|
|
|
// Convert task type string to TaskType
|
|
var taskType TaskType
|
|
switch result.TaskType {
|
|
case "feature_development":
|
|
taskType = TaskTypeFeatureDevelopment
|
|
case "bug_fix":
|
|
taskType = TaskTypeBugFix
|
|
case "refactoring":
|
|
taskType = TaskTypeRefactoring
|
|
case "security":
|
|
taskType = TaskTypeSecurity
|
|
case "integration":
|
|
taskType = TaskTypeIntegration
|
|
case "optimization":
|
|
taskType = TaskTypeOptimization
|
|
case "maintenance":
|
|
taskType = TaskTypeMaintenance
|
|
default:
|
|
taskType = TaskTypeFeatureDevelopment // Default fallback
|
|
}
|
|
|
|
classification := &TaskClassification{
|
|
TaskType: taskType,
|
|
ComplexityScore: result.ComplexityScore,
|
|
PrimaryDomains: result.PrimaryDomains,
|
|
SecondaryDomains: result.SecondaryDomains,
|
|
EstimatedDuration: result.EstimatedDuration,
|
|
RiskLevel: result.RiskLevel,
|
|
RequiredExperience: result.RequiredExperience,
|
|
}
|
|
|
|
return classification, nil
|
|
}
|
|
|
|
// ParseSkillRequirementsResponse parses LLM response into SkillRequirements
|
|
func (c *OllamaClient) ParseSkillRequirementsResponse(response string) (*SkillRequirements, error) {
|
|
// Clean the response - remove markdown code blocks if present
|
|
response = strings.TrimSpace(response)
|
|
response = strings.TrimPrefix(response, "```json")
|
|
response = strings.TrimPrefix(response, "```")
|
|
response = strings.TrimSuffix(response, "```")
|
|
response = strings.TrimSpace(response)
|
|
|
|
var result struct {
|
|
CriticalSkills []struct {
|
|
Domain string `json:"domain"`
|
|
MinProficiency float64 `json:"min_proficiency"`
|
|
Weight float64 `json:"weight"`
|
|
Critical bool `json:"critical"`
|
|
} `json:"critical_skills"`
|
|
DesirableSkills []struct {
|
|
Domain string `json:"domain"`
|
|
MinProficiency float64 `json:"min_proficiency"`
|
|
Weight float64 `json:"weight"`
|
|
Critical bool `json:"critical"`
|
|
} `json:"desirable_skills"`
|
|
TotalSkillCount int `json:"total_skill_count"`
|
|
}
|
|
|
|
if err := json.Unmarshal([]byte(response), &result); err != nil {
|
|
return nil, fmt.Errorf("failed to parse skill requirements response: %w", err)
|
|
}
|
|
|
|
// Convert to SkillRequirements format
|
|
critical := make([]SkillRequirement, len(result.CriticalSkills))
|
|
for i, skill := range result.CriticalSkills {
|
|
critical[i] = SkillRequirement{
|
|
Domain: skill.Domain,
|
|
MinProficiency: skill.MinProficiency,
|
|
Weight: skill.Weight,
|
|
Critical: skill.Critical,
|
|
}
|
|
}
|
|
|
|
desirable := make([]SkillRequirement, len(result.DesirableSkills))
|
|
for i, skill := range result.DesirableSkills {
|
|
desirable[i] = SkillRequirement{
|
|
Domain: skill.Domain,
|
|
MinProficiency: skill.MinProficiency,
|
|
Weight: skill.Weight,
|
|
Critical: skill.Critical,
|
|
}
|
|
}
|
|
|
|
skillRequirements := &SkillRequirements{
|
|
CriticalSkills: critical,
|
|
DesirableSkills: desirable,
|
|
TotalSkillCount: len(critical) + len(desirable),
|
|
}
|
|
|
|
return skillRequirements, nil
|
|
} |