Implement Phase 2: Core API Integration

- Add Claude API client with Anthropic SDK
  - Retry logic with exponential backoff (3 attempts)
  - Handle auth errors, rate limits, and server errors
  - Architecture classification with JSON parsing
  - Research query generation from architecture
- Add Perplexity API client with native fetch
  - Retry logic with exponential backoff
  - Parse responses with sources and citations
  - Execute multiple research queries
- Add prompt templates for architecture classification
- Add default query generation as fallback
- Add ArchitectureGenerator combining Claude + Perplexity
- Add 44 new tests (80 total, all passing)

All Phase 2 acceptance criteria met:
- Claude client sends prompts and retries on failure
- Perplexity client searches and parses with sources
- Architecture classification returns valid JSON
- Research queries generated (4-6 per architecture)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

.claude/ralph-loop.local.md

@@ -1,37 +1,40 @@
 ---
 active: true
 iteration: 1
-max_iterations: 30
-completion_promise: "PHASE_1_COMPLETE"
-started_at: "2026-01-10T12:00:45Z"
+max_iterations: 40
+completion_promise: "PHASE_2_COMPLETE"
+started_at: "2026-01-10T12:10:57Z"
 ---
 
-Execute Phase 1 - Foundation from PROMPT.md.
+Execute Phase 2 - Core API Integration from PROMPT.md.
 
-You are building the Ralph PRD Generator CLI tool.
+Phase 1 is complete. Now implement Phase 2.
 
-Read PROMPT.md for full specification.
-Read prd.json for feature tracking.
-Read CLAUDE.md for project configuration.
+Read PROMPT.md for API specifications.
+Read prd.json for Phase 2 features.
 
-Phase 1 tasks:
-1. Set up TypeScript project with all dependencies
-2. Configure tsup, vitest, eslint
-3. Implement CLI framework with Commander.js
-4. Implement config management for API key storage
-5. Create basic project structure
-6. Write tests for config and CLI
+Phase 2 tasks:
+1. Claude API client with Anthropic SDK
+2. Perplexity API client with fetch
+3. Architecture classification prompt and parser
+4. Research query generation
+5. Tests for all API clients
+
+API Integration notes:
+- Use environment variables for API keys during testing
+- Implement retry logic with 3 attempts and exponential backoff
+- Handle rate limits gracefully
+- Parse JSON responses safely
 
 For each feature:
-1. Write tests first
+1. Write tests first using mocked API calls
 2. Implement the feature
-3. Run: npm run build && npm run test && npm run lint
-4. If all pass, update prd.json by setting passes to true for completed features
-5. Commit with descriptive message
-6. Append progress to progress.txt
+3. Verify: npm run build && npm run test && npm run lint
+4. Update prd.json when feature passes
+5. Commit and log progress
 
-When ALL Phase 1 features in prd.json have passes set to true:
-Output <promise>PHASE_1_COMPLETE</promise>
+When ALL Phase 2 features pass:
+Output <promise>PHASE_2_COMPLETE</promise>
 
-If blocked after 10 attempts on same issue:
-Document in progress.txt and output <promise>PHASE_1_BLOCKED</promise>
+If blocked:
+Output <promise>PHASE_2_BLOCKED</promise>