runs - review pending

2026-01-27 13:58:54 -05:00
parent c85b877dc0
commit 7d60504e1d
10 changed files with 779 additions and 110 deletions
--- a/.env
+++ b/.env
@@ -8,7 +8,14 @@ REDIS_URL=redis://redis:6379
 MEILISEARCH_URL=http://meilisearch:7700
 MEILISEARCH_API_KEY=your_secure_master_key_here

-# OpenAI
+# AI Provider Configuration
+# Option 1: Ollama (Local LLM - DEFAULT, no API costs!)
+USE_OLLAMA=true
+OLLAMA_URL=http://ollama:11434
+OLLAMA_MODEL=gpt-oss:latest
+
+# Option 2: OpenAI (Cloud - optional fallback)
+# Set these if you want OpenAI as fallback when Ollama is unavailable
 OPENAI_API_KEY=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
 OPENAI_MODEL=gpt-4o-mini

--- a/.env.example
+++ b/.env.example
@@ -8,9 +8,16 @@ REDIS_URL=redis://redis:6379
 MEILISEARCH_URL=http://meilisearch:7700
 MEILISEARCH_API_KEY=your_secure_master_key_here

-# OpenAI
-OPENAI_API_KEY=sk-your-openai-api-key-here
-OPENAI_MODEL=gpt-4o
+# AI Provider Configuration
+# Option 1: Ollama (Local LLM - DEFAULT, no API costs!)
+USE_OLLAMA=true
+OLLAMA_URL=http://ollama:11434
+OLLAMA_MODEL=gpt-oss:latest
+
+# Option 2: OpenAI (Cloud - requires API key, falls back if Ollama unavailable)
+# Uncomment and set these if you want to use OpenAI as fallback
+# OPENAI_API_KEY=sk-your-openai-api-key-here
+# OPENAI_MODEL=gpt-4o

 # Admin Credentials (change these!)
 ADMIN_USERNAME=admin
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -18,7 +18,7 @@ This file provides essential context and guidelines for AI agents working on thi
 - **Search**: Meilisearch v1.6
 - **Cache**: Redis 7
 - **Templating**: EJS
- **AI**: OpenAI API (GPT-4o/GPT-4-turbo)
+- **AI**: Ollama (local LLM - gpt-oss:latest) with OpenAI fallback
 - **Containerization**: Docker + Docker Compose
 - **Hosting**: Self-hosted on Linode

--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -14,6 +14,9 @@ services:
      - MEILISEARCH_API_KEY=${MEILISEARCH_API_KEY:-master_key}
      - OPENAI_API_KEY=${OPENAI_API_KEY}
      - OPENAI_MODEL=${OPENAI_MODEL:-gpt-4o}
+      - OLLAMA_URL=http://ollama:11434
+      - OLLAMA_MODEL=${OLLAMA_MODEL:-gpt-oss:latest}
+      - USE_OLLAMA=${USE_OLLAMA:-true}
      - ADMIN_USERNAME=${ADMIN_USERNAME:-admin}
      - ADMIN_PASSWORD=${ADMIN_PASSWORD:-changeme}
      - SESSION_SECRET=${SESSION_SECRET:-changeme}
@@ -25,6 +28,8 @@ services:
        condition: service_started
      meilisearch:
        condition: service_started
+      ollama:
+        condition: service_started
    volumes:
      - ./src:/app/src
      - ./migrations:/app/migrations
@@ -33,6 +38,35 @@ services:
    networks:
      - privacy-analyzer-network

+  ollama:
+    image: ollama/ollama:latest
+    container_name: privacy-analyzer-ollama
+    ports:
+      - "11434:11434"
+    volumes:
+      - ollama_data:/root/.ollama
+    environment:
+      - OLLAMA_ORIGINS=*
+      - OLLAMA_HOST=0.0.0.0
+    entrypoint: >
+      sh -c "
+        ollama serve &
+        sleep 5
+        echo 'Pulling ${OLLAMA_MODEL:-gpt-oss:latest} model...'
+        ollama pull ${OLLAMA_MODEL:-gpt-oss:latest}
+        echo 'Model ready!'
+        wait
+      "
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
+      interval: 30s
+      timeout: 10s
+      retries: 5
+      start_period: 60s
+    restart: unless-stopped
+    networks:
+      - privacy-analyzer-network
+
  postgres:
    image: postgres:15-alpine
    container_name: privacy-analyzer-postgres
@@ -86,6 +120,8 @@ volumes:
    driver: local
  meilisearch_data:
    driver: local
+  ollama_data:
+    driver: local

 networks:
  privacy-analyzer-network:
--- a/src/app.js
+++ b/src/app.js
@@ -6,6 +6,8 @@ import { PolicyFetcher } from './services/policyFetcher.js';
 import { AIAnalyzer } from './services/aiAnalyzer.js';
 import { Scheduler } from './services/scheduler.js';
 import { SearchIndexer } from './services/searchIndexer.js';
+import { AnalysisQueue } from './services/analysisQueue.js';
+import { AnalysisWorker } from './services/analysisWorker.js';
 import ejs from 'ejs';
 import { readFile } from 'fs/promises';
 import { join, dirname } from 'path';
@@ -258,6 +260,10 @@ async function handleRequest(req) {
        !s.last_analyzed || new Date(s.last_analyzed) < thirtyDaysAgo
      ).length;
      
+      // Get flash message from query params
+      const message = url.searchParams.get('message');
+      const jobId = url.searchParams.get('jobId');
+      
      const html = await renderTemplate('admin/dashboard', {
        title: 'Admin Dashboard',
        description: 'Manage services and privacy analyses',
@@ -267,7 +273,9 @@ async function handleRequest(req) {
          totalServices: services.length,
          totalAnalyses,
          pendingUpdates
-        }
+        },
+        message,
+        jobId
      });
      
      return new Response(html, {
@@ -478,49 +486,83 @@ async function handleRequest(req) {
          return new Response('Service not found', { status: 404 });
        }
        
-        console.log(`Starting analysis for service: ${service.name}`);
+        console.log(`Queueing analysis for service: ${service.name}`);
        
-        // Fetch policy
-        const policyData = await PolicyFetcher.fetchPolicy(service.policy_url);
+        // Add to queue instead of processing synchronously
+        const jobId = await AnalysisQueue.addJob(id, service.name);
        
-        // Check if content has changed
-        const contentHash = PolicyVersion.generateContentHash(policyData.content);
-        const existingVersion = await PolicyVersion.findByContentHash(id, contentHash);
+        // Redirect to dashboard with success message
+        return new Response(null, { 
+          status: 302, 
+          headers: { 
+            Location: `/admin/dashboard?message=Analysis queued for ${encodeURIComponent(service.name)}&jobId=${jobId}` 
+          } 
+        });
+      } catch (error) {
+        console.error('Queue error:', error);
+        return new Response(`Failed to queue analysis: ${error.message}`, { status: 500 });
+      }
+    }
    
-        let policyVersion;
-        if (existingVersion) {
-          console.log('Policy content unchanged, using existing version');
-          policyVersion = existingVersion;
-        } else {
-          console.log('New policy content detected, creating new version');
-          policyVersion = await PolicyVersion.create({
-            service_id: id,
-            content: policyData.content,
-            fetched_at: policyData.fetchedAt
+    // Get analysis job status - GET /api/analysis/status/:jobId
+    if (method === 'GET' && pathname.match(/^\/api\/analysis\/status\/[^\/]+$/)) {
+      try {
+        const match = pathname.match(/^\/api\/analysis\/status\/([^\/]+)$/);
+        const jobId = match[1];
+        
+        const jobStatus = await AnalysisQueue.getJobStatus(jobId);
+        
+        if (!jobStatus) {
+          return new Response(JSON.stringify({ error: 'Job not found' }), {
+            status: 404,
+            headers: { 'Content-Type': 'application/json' }
          });
        }
        
-        // Analyze with AI
-        const analysisResult = await AIAnalyzer.analyzePolicy(policyData.content);
-        
-        // Save analysis
-        const analysis = await Analysis.create({
-          service_id: id,
-          policy_version_id: policyVersion.id,
-          overall_score: analysisResult.overall_score,
-          findings: analysisResult.findings,
-          raw_analysis: analysisResult.raw_response
-        });
-        
-        console.log(`Analysis complete for ${service.name}: Grade ${analysis.overall_score}`);
-        
-        return new Response(null, { 
-          status: 302, 
-          headers: { Location: '/admin/dashboard' } 
+        return new Response(JSON.stringify(jobStatus), {
+          headers: { 'Content-Type': 'application/json' }
        });
      } catch (error) {
-        console.error('Analysis error:', error);
-        return new Response(`Analysis failed: ${error.message}`, { status: 500 });
+        console.error('Status check error:', error);
+        return new Response(JSON.stringify({ error: error.message }), {
+          status: 500,
+          headers: { 'Content-Type': 'application/json' }
+        });
+      }
+    }
+    
+    // Get queue status - GET /api/analysis/queue
+    if (method === 'GET' && pathname === '/api/analysis/queue') {
+      const sessionToken = req.cookies?.session_token;
+      
+      if (!sessionToken || !(await AdminSession.findByToken(sessionToken))) {
+        return new Response(JSON.stringify({ error: 'Unauthorized' }), {
+          status: 401,
+          headers: { 'Content-Type': 'application/json' }
+        });
+      }
+      
+      try {
+        const [queueLength, pendingJobs, processingJobs] = await Promise.all([
+          AnalysisQueue.getQueueLength(),
+          AnalysisQueue.getPendingJobs(),
+          AnalysisQueue.getProcessingJobs()
+        ]);
+        
+        return new Response(JSON.stringify({
+          queueLength,
+          pending: pendingJobs,
+          processing: processingJobs,
+          workerStatus: AnalysisWorker.getStatus()
+        }), {
+          headers: { 'Content-Type': 'application/json' }
+        });
+      } catch (error) {
+        console.error('Queue status error:', error);
+        return new Response(JSON.stringify({ error: error.message }), {
+          status: 500,
+          headers: { 'Content-Type': 'application/json' }
+        });
      }
    }
    
@@ -585,7 +627,7 @@ const server = Bun.serve({

 console.log(`Server running at http://localhost:${server.port}`);

-// Initialize scheduler and search indexer (non-blocking)
+// Initialize scheduler, search indexer, and analysis worker (non-blocking)
 (async () => {
  try {
    // Initialize scheduler
@@ -594,6 +636,10 @@ console.log(`Server running at http://localhost:${server.port}`);
    // Initialize search indexer
    await SearchIndexer.init();
    console.log('Search indexer initialized');
+    
+    // Start analysis worker
+    AnalysisWorker.start();
+    console.log('Analysis worker started');
  } catch (error) {
    console.error('Failed to initialize services:', error);
  }
--- a/src/scripts/test-ollama.js
+++ b/src/scripts/test-ollama.js
@@ -0,0 +1,66 @@
+import { AIAnalyzer } from '../services/aiAnalyzer.js';
+
+async function testOllama() {
+  console.log('Testing Ollama integration...');
+  console.log('Ollama URL:', AIAnalyzer.OLLAMA_URL);
+  console.log('Ollama Model:', AIAnalyzer.OLLAMA_MODEL);
+  
+  // Check if Ollama is available
+  const isAvailable = await AIAnalyzer.isOllamaAvailable();
+  console.log('Ollama available:', isAvailable);
+  
+  if (!isAvailable) {
+    console.log('Ollama is not available. Make sure the ollama container is running.');
+    console.log('Run: docker-compose up -d ollama');
+    process.exit(1);
+  }
+  
+  const testPolicy = `Privacy Policy
+
+Information We Collect:
+- Name and email address when you create an account
+- Usage data including pages visited and features used
+- Device information such as browser type and IP address
+
+How We Use Your Information:
+- To provide and improve our services
+- To communicate with you about updates and features
+- For analytics and service optimization
+
+Data Sharing:
+We share data with Google Analytics for usage analytics.
+
+Your Rights:
+- You can access and update your personal information
+- You can request deletion of your account and data
+- You can opt out of marketing communications
+
+Cookies:
+We use cookies for session management and analytics.
+
+Data Retention:
+We retain your data for as long as your account is active.`;
+
+  try {
+    console.log('\nSending test policy to Ollama...');
+    const result = await AIAnalyzer.analyzeWithOllama(testPolicy);
+    
+    console.log('\n✓ Ollama analysis successful!');
+    console.log('\nAnalysis Results:');
+    console.log('Overall Score:', result.overall_score);
+    console.log('Summary:', result.summary);
+    console.log('\nScore Breakdown:');
+    console.log(JSON.stringify(result.score_breakdown, null, 2));
+    console.log('\nFindings:');
+    console.log('Positive:', result.findings.positive.length);
+    console.log('Negative:', result.findings.negative.length);
+    console.log('Neutral:', result.findings.neutral.length);
+    
+  } catch (error) {
+    console.error('✗ Test failed:', error.message);
+    console.error('Stack:', error.stack);
+    process.exit(1);
+  }
+}
+
+testOllama();
--- a/src/services/aiAnalyzer.js
+++ b/src/services/aiAnalyzer.js
@@ -1,5 +1,6 @@
 /**
- * Service to analyze privacy policies using OpenAI
+ * Service to analyze privacy policies using Ollama (local LLM)
+ * Falls back to OpenAI if Ollama is not available
 */

 import openai from '../config/openai.js';
@@ -7,6 +8,9 @@ import openai from '../config/openai.js';
 export class AIAnalyzer {
  static MAX_RETRIES = 3;
  static RETRY_DELAY_MS = 2000;
+  static OLLAMA_URL = process.env.OLLAMA_URL || 'http://ollama:11434';
+  static OLLAMA_MODEL = process.env.OLLAMA_MODEL || 'gpt-oss:latest';
+  static USE_OLLAMA = process.env.USE_OLLAMA === 'true' || true; // Default to Ollama

  /**
   * System prompt for the AI
@@ -78,7 +82,128 @@ Important:
 - Mark critical issues as "blocker", significant issues as "bad"
 - Include at least 2-3 positive aspects if they exist
 - Include all significant negative aspects
- Be objective and factual based on the policy text provided`;
+- Be objective and factual based on the policy text provided
+- Respond ONLY with valid JSON, no markdown formatting, no code blocks`;
+
+  /**
+   * Check if Ollama is available
+   * @returns {Promise<boolean>}
+   */
+  static async isOllamaAvailable() {
+    try {
+      const response = await fetch(`${this.OLLAMA_URL}/api/tags`, {
+        method: 'GET',
+        timeout: 5000
+      });
+      return response.ok;
+    } catch (error) {
+      console.log('Ollama not available:', error.message);
+      return false;
+    }
+  }
+
+  /**
+   * Analyze using Ollama
+   * @param {string} policyText - Policy text
+   * @returns {Promise<Object>}
+   */
+  static async analyzeWithOllama(policyText) {
+    console.log(`Analyzing with Ollama (${this.OLLAMA_MODEL})...`);
+    
+    const response = await fetch(`${this.OLLAMA_URL}/api/generate`, {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json'
+      },
+      body: JSON.stringify({
+        model: this.OLLAMA_MODEL,
+        prompt: `${this.SYSTEM_PROMPT}\n\nAnalyze this privacy policy and respond with ONLY valid JSON:\n\n${policyText}`,
+        stream: false,
+        options: {
+          temperature: 0.3,
+          num_predict: 4000
+        }
+      })
+    });
+
+    if (!response.ok) {
+      throw new Error(`Ollama error: ${response.status} ${response.statusText}`);
+    }
+
+    const result = await response.json();
+    
+    // Parse the response - Ollama returns the full response in 'response' field
+    let content = result.response;
+    
+    // Try to extract JSON from the response (in case there's extra text)
+    const jsonMatch = content.match(/\{[\s\S]*\}/);
+    if (jsonMatch) {
+      content = jsonMatch[0];
+    }
+    
+    let analysis;
+    try {
+      analysis = JSON.parse(content);
+    } catch (parseError) {
+      console.error('Failed to parse Ollama response:', content);
+      throw new Error('Invalid JSON response from Ollama');
+    }
+
+    this.validateAnalysis(analysis);
+
+    return {
+      ...analysis,
+      raw_response: result.response,
+      model: this.OLLAMA_MODEL,
+      provider: 'ollama'
+    };
+  }
+
+  /**
+   * Analyze using OpenAI
+   * @param {string} policyText - Policy text
+   * @returns {Promise<Object>}
+   */
+  static async analyzeWithOpenAI(policyText) {
+    console.log('Analyzing with OpenAI...');
+    
+    const response = await openai.chat.completions.create({
+      model: process.env.OPENAI_MODEL || 'gpt-4o',
+      messages: [
+        {
+          role: 'system',
+          content: this.SYSTEM_PROMPT
+        },
+        {
+          role: 'user',
+          content: `Analyze this privacy policy and provide a structured assessment:\n\n${policyText}`
+        }
+      ],
+      response_format: { type: 'json_object' },
+      temperature: 0.3,
+      max_tokens: 4000
+    });
+
+    const content = response.choices[0].message.content;
+    
+    let analysis;
+    try {
+      analysis = JSON.parse(content);
+    } catch (parseError) {
+      console.error('Failed to parse OpenAI response:', parseError);
+      throw new Error('Invalid JSON response from OpenAI');
+    }
+
+    this.validateAnalysis(analysis);
+
+    return {
+      ...analysis,
+      raw_response: content,
+      model: response.model,
+      usage: response.usage,
+      provider: 'openai'
+    };
+  }

  /**
   * Analyze a privacy policy
@@ -86,13 +211,12 @@ Important:
   * @returns {Promise<Object>} - Analysis results
   */
  static async analyzePolicy(policyText) {
-    // Token-efficient truncation - reduce to ~8000 chars (~2000 tokens) for cost savings
+    // Token-efficient truncation
    const maxLength = 8000;
    let truncatedText = policyText;
    
    if (policyText.length > maxLength) {
      console.warn(`Policy text too long (${policyText.length} chars), truncating to ${maxLength} to save tokens`);
-      // Keep the beginning (usually contains important info) and end (usually has rights/legal)
      const beginning = policyText.substring(0, Math.floor(maxLength * 0.7));
      const end = policyText.substring(policyText.length - Math.floor(maxLength * 0.25));
      truncatedText = beginning + '\n\n[... Content truncated to save tokens ...]\n\n' + end;
@@ -100,59 +224,36 @@ Important:

    let lastError;
    
+    // Try Ollama first if enabled
+    if (this.USE_OLLAMA) {
+      try {
+        const isAvailable = await this.isOllamaAvailable();
+        if (isAvailable) {
+          console.log('Ollama is available, using local model');
+          return await this.analyzeWithOllama(truncatedText);
+        }
+      } catch (error) {
+        console.log('Ollama analysis failed, falling back to OpenAI:', error.message);
+      }
+    }
+
+    // Fallback to OpenAI
    for (let attempt = 1; attempt <= this.MAX_RETRIES; attempt++) {
      try {
-        console.log(`Analyzing policy with OpenAI (attempt ${attempt}/${this.MAX_RETRIES})`);
+        console.log(`Analyzing with OpenAI (attempt ${attempt}/${this.MAX_RETRIES})`);
        
-        const response = await openai.chat.completions.create({
-          model: process.env.OPENAI_MODEL || 'gpt-4o',
-          messages: [
-            {
-              role: 'system',
-              content: this.SYSTEM_PROMPT
-            },
-            {
-              role: 'user',
-              content: `Analyze this privacy policy and provide a structured assessment:\n\n${truncatedText}`
-            }
-          ],
-          response_format: { type: 'json_object' },
-          temperature: 0.3, // Lower temperature for more consistent results
-          max_tokens: 4000
-        });
+        const result = await this.analyzeWithOpenAI(truncatedText);
        
-        const content = response.choices[0].message.content;
-        
-        // Parse the JSON response
-        let analysis;
-        try {
-          analysis = JSON.parse(content);
-        } catch (parseError) {
-          console.error('Failed to parse AI response as JSON:', parseError);
-          throw new Error('Invalid JSON response from AI');
-        }
-
-        // Validate response structure
-        this.validateAnalysis(analysis);
-
-        console.log(`Analysis complete. Overall score: ${analysis.overall_score}`);
-        
-        return {
-          ...analysis,
-          raw_response: content,
-          model: response.model,
-          usage: response.usage
-        };
+        console.log(`Analysis complete. Overall score: ${result.overall_score}`);
+        return result;

      } catch (error) {
        lastError = error;
        console.error(`Analysis attempt ${attempt} failed:`, error.message);
        
-        // Check for specific error types
        if (error.status === 429 || error.message?.includes('429')) {
          const errorMsg = 'OpenAI API rate limit exceeded. Please check your plan and billing details at https://platform.openai.com/account/billing';
          console.error(errorMsg);
-          // Don't retry on rate limit - it won't help
          throw new Error(errorMsg);
        }
        
@@ -182,19 +283,16 @@ Important:
    const validGrades = ['A', 'B', 'C', 'D', 'E'];
    const requiredFields = ['overall_score', 'score_breakdown', 'findings', 'summary'];
    
-    // Check required fields
    for (const field of requiredFields) {
      if (!(field in analysis)) {
        throw new Error(`Missing required field: ${field}`);
      }
    }
    
-    // Validate overall_score
    if (!validGrades.includes(analysis.overall_score)) {
      throw new Error(`Invalid overall_score: ${analysis.overall_score}`);
    }
    
-    // Validate score_breakdown
    const breakdownFields = ['data_collection', 'data_sharing', 'user_rights', 'data_retention', 'tracking_security'];
    for (const field of breakdownFields) {
      if (!(field in analysis.score_breakdown)) {
@@ -205,7 +303,6 @@ Important:
      }
    }
    
-    // Validate findings
    if (!Array.isArray(analysis.findings.positive)) {
      analysis.findings.positive = [];
    }
@@ -216,7 +313,6 @@ Important:
      analysis.findings.neutral = [];
    }
    
-    // Ensure arrays
    if (!Array.isArray(analysis.data_types_collected)) {
      analysis.data_types_collected = [];
    }
@@ -225,24 +321,6 @@ Important:
    }
  }

-  /**
-   * Calculate overall score from breakdown
-   * @param {Object} breakdown - Score breakdown
-   * @returns {string} - Overall grade
-   */
-  static calculateOverallScore(breakdown) {
-    const gradeValues = { A: 5, B: 4, C: 3, D: 2, E: 1 };
-    const grades = Object.values(breakdown);
-    const total = grades.reduce((sum, grade) => sum + (gradeValues[grade] || 0), 0);
-    const average = total / grades.length;
-    
-    if (average >= 4.5) return 'A';
-    if (average >= 3.5) return 'B';
-    if (average >= 2.5) return 'C';
-    if (average >= 1.5) return 'D';
-    return 'E';
-  }
-
  /**
   * Sleep helper
   * @param {number} ms - Milliseconds to sleep
--- a/src/services/analysisQueue.js
+++ b/src/services/analysisQueue.js
@@ -0,0 +1,178 @@
+/**
+ * Redis-based job queue for background processing
+ */
+
+import redis from '../config/redis.js';
+
+const QUEUE_KEY = 'analysis:queue';
+const PROCESSING_KEY = 'analysis:processing';
+const RESULTS_KEY_PREFIX = 'analysis:result:';
+const RESULTS_TTL = 3600; // 1 hour
+
+export class AnalysisQueue {
+  /**
+   * Add analysis job to queue
+   * @param {number} serviceId - Service ID
+   * @param {string} serviceName - Service name for display
+   * @returns {Promise<string>} - Job ID
+   */
+  static async addJob(serviceId, serviceName) {
+    const jobId = `job-${Date.now()}-${serviceId}`;
+    const job = {
+      id: jobId,
+      serviceId,
+      serviceName,
+      status: 'pending',
+      createdAt: new Date().toISOString(),
+      updatedAt: new Date().toISOString()
+    };
+    
+    // Add to queue
+    await redis.lpush(QUEUE_KEY, JSON.stringify(job));
+    
+    console.log(`Added analysis job ${jobId} for service ${serviceName}`);
+    return jobId;
+  }
+
+  /**
+   * Get next job from queue
+   * @returns {Promise<Object|null>} - Job or null if queue empty
+   */
+  static async getNextJob() {
+    // Pop from the end of the queue (FIFO)
+    const jobData = await redis.rpop(QUEUE_KEY);
+    
+    if (!jobData) {
+      return null;
+    }
+    
+    const job = JSON.parse(jobData);
+    job.status = 'processing';
+    job.updatedAt = new Date().toISOString();
+    
+    // Add to processing set
+    await redis.hset(PROCESSING_KEY, job.id, JSON.stringify(job));
+    
+    return job;
+  }
+
+  /**
+   * Mark job as completed
+   * @param {string} jobId - Job ID
+   * @param {Object} result - Analysis result
+   */
+  static async completeJob(jobId, result) {
+    const jobData = await redis.hget(PROCESSING_KEY, jobId);
+    
+    if (!jobData) {
+      console.warn(`Job ${jobId} not found in processing set`);
+      return;
+    }
+    
+    const job = JSON.parse(jobData);
+    job.status = 'completed';
+    job.result = result;
+    job.completedAt = new Date().toISOString();
+    job.updatedAt = new Date().toISOString();
+    
+    // Remove from processing
+    await redis.hdel(PROCESSING_KEY, jobId);
+    
+    // Store result with TTL
+    await redis.setex(
+      `${RESULTS_KEY_PREFIX}${jobId}`,
+      RESULTS_TTL,
+      JSON.stringify(job)
+    );
+    
+    console.log(`Completed analysis job ${jobId}`);
+  }
+
+  /**
+   * Mark job as failed
+   * @param {string} jobId - Job ID
+   * @param {string} error - Error message
+   */
+  static async failJob(jobId, error) {
+    const jobData = await redis.hget(PROCESSING_KEY, jobId);
+    
+    if (!jobData) {
+      console.warn(`Job ${jobId} not found in processing set`);
+      return;
+    }
+    
+    const job = JSON.parse(jobData);
+    job.status = 'failed';
+    job.error = error;
+    job.failedAt = new Date().toISOString();
+    job.updatedAt = new Date().toISOString();
+    
+    // Remove from processing
+    await redis.hdel(PROCESSING_KEY, jobId);
+    
+    // Store failed result with TTL
+    await redis.setex(
+      `${RESULTS_KEY_PREFIX}${jobId}`,
+      RESULTS_TTL,
+      JSON.stringify(job)
+    );
+    
+    console.error(`Failed analysis job ${jobId}: ${error}`);
+  }
+
+  /**
+   * Get job status
+   * @param {string} jobId - Job ID
+   * @returns {Promise<Object|null>}
+   */
+  static async getJobStatus(jobId) {
+    // Check results first
+    const result = await redis.get(`${RESULTS_KEY_PREFIX}${jobId}`);
+    if (result) {
+      return JSON.parse(result);
+    }
+    
+    // Check processing
+    const processing = await redis.hget(PROCESSING_KEY, jobId);
+    if (processing) {
+      return JSON.parse(processing);
+    }
+    
+    return null;
+  }
+
+  /**
+   * Get queue length
+   * @returns {Promise<number>}
+   */
+  static async getQueueLength() {
+    return await redis.llen(QUEUE_KEY);
+  }
+
+  /**
+   * Get all pending jobs
+   * @returns {Promise<Array>}
+   */
+  static async getPendingJobs() {
+    const jobs = await redis.lrange(QUEUE_KEY, 0, -1);
+    return jobs.map(j => JSON.parse(j));
+  }
+
+  /**
+   * Get all processing jobs
+   * @returns {Promise<Array>}
+   */
+  static async getProcessingJobs() {
+    const jobs = await redis.hgetall(PROCESSING_KEY);
+    return Object.values(jobs).map(j => JSON.parse(j));
+  }
+
+  /**
+   * Clear all queues (use with caution)
+   */
+  static async clearAll() {
+    await redis.del(QUEUE_KEY);
+    await redis.del(PROCESSING_KEY);
+    console.log('Cleared all analysis queues');
+  }
+}
--- a/src/services/analysisWorker.js
+++ b/src/services/analysisWorker.js
@@ -0,0 +1,151 @@
+/**
+ * Background worker for processing analysis jobs
+ */
+
+import { AnalysisQueue } from './analysisQueue.js';
+import { Service } from '../models/Service.js';
+import { PolicyVersion } from '../models/PolicyVersion.js';
+import { Analysis } from '../models/Analysis.js';
+import { PolicyFetcher } from './policyFetcher.js';
+import { AIAnalyzer } from './aiAnalyzer.js';
+
+export class AnalysisWorker {
+  static isRunning = false;
+  static currentJob = null;
+  static pollInterval = 5000; // Check every 5 seconds
+
+  /**
+   * Start the worker
+   */
+  static start() {
+    if (this.isRunning) {
+      console.log('Analysis worker already running');
+      return;
+    }
+
+    this.isRunning = true;
+    console.log('Starting analysis worker...');
+    
+    this.processLoop();
+  }
+
+  /**
+   * Stop the worker
+   */
+  static stop() {
+    this.isRunning = false;
+    console.log('Analysis worker stopped');
+  }
+
+  /**
+   * Main processing loop
+   */
+  static async processLoop() {
+    while (this.isRunning) {
+      try {
+        // Try to get a job
+        const job = await AnalysisQueue.getNextJob();
+        
+        if (job) {
+          this.currentJob = job;
+          console.log(`Processing job ${job.id} for ${job.serviceName}`);
+          
+          try {
+            await this.processJob(job);
+          } catch (error) {
+            console.error(`Job ${job.id} failed:`, error);
+            await AnalysisQueue.failJob(job.id, error.message);
+          }
+          
+          this.currentJob = null;
+        } else {
+          // No jobs, wait before checking again
+          await this.sleep(this.pollInterval);
+        }
+      } catch (error) {
+        console.error('Worker error:', error);
+        await this.sleep(this.pollInterval);
+      }
+    }
+  }
+
+  /**
+   * Process a single job
+   * @param {Object} job - Job data
+   */
+  static async processJob(job) {
+    const { serviceId, id: jobId } = job;
+    
+    // Get service details
+    const service = await Service.findById(serviceId);
+    if (!service) {
+      throw new Error(`Service ${serviceId} not found`);
+    }
+
+    console.log(`[${jobId}] Fetching policy for ${service.name}...`);
+    
+    // Fetch policy
+    const policyData = await PolicyFetcher.fetchPolicy(service.policy_url);
+    
+    // Check if content has changed
+    const contentHash = PolicyVersion.generateContentHash(policyData.content);
+    const existingVersion = await PolicyVersion.findByContentHash(serviceId, contentHash);
+    
+    let policyVersion;
+    if (existingVersion) {
+      console.log(`[${jobId}] Policy unchanged, using existing version`);
+      policyVersion = existingVersion;
+    } else {
+      console.log(`[${jobId}] New policy version detected`);
+      policyVersion = await PolicyVersion.create({
+        service_id: serviceId,
+        content: policyData.content,
+        fetched_at: policyData.fetchedAt
+      });
+    }
+
+    console.log(`[${jobId}] Analyzing with AI...`);
+    
+    // Analyze with AI
+    const analysisResult = await AIAnalyzer.analyzePolicy(policyData.content);
+    
+    console.log(`[${jobId}] Saving analysis results...`);
+    
+    // Save analysis
+    const analysis = await Analysis.create({
+      service_id: serviceId,
+      policy_version_id: policyVersion.id,
+      overall_score: analysisResult.overall_score,
+      findings: analysisResult.findings,
+      raw_analysis: analysisResult.raw_response
+    });
+
+    console.log(`[${jobId}] Analysis complete: Grade ${analysis.overall_score}`);
+    
+    // Mark job as complete
+    await AnalysisQueue.completeJob(jobId, {
+      analysisId: analysis.id,
+      overallScore: analysis.overall_score,
+      model: analysisResult.model || analysisResult.provider,
+      contentLength: policyData.content.length
+    });
+  }
+
+  /**
+   * Get worker status
+   */
+  static getStatus() {
+    return {
+      isRunning: this.isRunning,
+      currentJob: this.currentJob,
+      queueLength: null // Would need async call
+    };
+  }
+
+  /**
+   * Sleep helper
+   */
+  static sleep(ms) {
+    return new Promise(resolve => setTimeout(resolve, ms));
+  }
+}
--- a/src/views/admin/dashboard.ejs
+++ b/src/views/admin/dashboard.ejs
@@ -1,6 +1,20 @@
 <div class="admin-dashboard">
    <h1>Admin Dashboard</h1>
    
+    <% if (typeof message !== 'undefined' && message) { %>
+        <div class="alert alert-success" id="flash-message">
+            <%= message %>
+        </div>
+    <% } %>
+    
+    <div id="analysis-queue-status" class="queue-status" style="display: none;">
+        <div class="queue-info">
+            <span class="queue-badge" id="queue-count">0</span>
+            <span>analyses in queue</span>
+        </div>
+        <div id="active-jobs"></div>
+    </div>
+    
    <div class="stats-grid">
        <div class="stat-card">
            <h3>Total Services</h3>
@@ -159,4 +173,90 @@
        padding: 3rem;
        color: #666;
    }
+    
+    .alert {
+        padding: 1rem;
+        margin-bottom: 1rem;
+        border-radius: 4px;
+    }
+    
+    .alert-success {
+        background: #d4edda;
+        color: #155724;
+        border: 1px solid #c3e6cb;
+    }
+    
+    .queue-status {
+        background: #fff;
+        padding: 1rem;
+        border-radius: 8px;
+        margin-bottom: 1rem;
+        box-shadow: 0 2px 4px rgba(0,0,0,0.1);
+    }
+    
+    .queue-info {
+        display: flex;
+        align-items: center;
+        gap: 0.5rem;
+        margin-bottom: 0.5rem;
+    }
+    
+    .queue-badge {
+        background: #3498db;
+        color: white;
+        padding: 0.25rem 0.75rem;
+        border-radius: 12px;
+        font-weight: bold;
+    }
+    
+    .processing-indicator {
+        color: #e67e22;
+        font-style: italic;
+    }
 </style>
+
+<script>
+    // Poll for queue status every 5 seconds
+    async function updateQueueStatus() {
+        try {
+            const response = await fetch('/api/analysis/queue');
+            if (!response.ok) return;
+            
+            const data = await response.json();
+            
+            const queueStatus = document.getElementById('analysis-queue-status');
+            const queueCount = document.getElementById('queue-count');
+            const activeJobs = document.getElementById('active-jobs');
+            
+            if (data.queueLength > 0 || data.processing.length > 0) {
+                queueStatus.style.display = 'block';
+                queueCount.textContent = data.queueLength;
+                
+                if (data.processing.length > 0) {
+                    activeJobs.innerHTML = '<div class="processing-indicator">⚙️ Processing: ' + 
+                        data.processing.map(j => j.serviceName).join(', ') + '</div>';
+                } else {
+                    activeJobs.innerHTML = '';
+                }
+            } else {
+                queueStatus.style.display = 'none';
+            }
+        } catch (error) {
+            console.error('Failed to update queue status:', error);
+        }
+    }
+    
+    // Update immediately and then every 5 seconds
+    updateQueueStatus();
+    setInterval(updateQueueStatus, 5000);
+    
+    // Auto-hide flash message after 5 seconds
+    const flashMessage = document.getElementById('flash-message');
+    if (flashMessage) {
+        setTimeout(() => {
+            flashMessage.style.opacity = '0';
+            flashMessage.style.transition = 'opacity 0.5s';
+            setTimeout(() => flashMessage.remove(), 500);
+        }, 5000);
+    }
+</script>