tasq/node_modules/agentic-flow/docs/plans/requesty
2026-04-09 19:01:53 +08:00
..
01-api-research.md added ruflo 2026-04-09 19:01:53 +08:00
02-architecture.md added ruflo 2026-04-09 19:01:53 +08:00
03-implementation-phases.md added ruflo 2026-04-09 19:01:53 +08:00
04-testing-strategy.md added ruflo 2026-04-09 19:01:53 +08:00
05-migration-guide.md added ruflo 2026-04-09 19:01:53 +08:00
00-overview.md added ruflo 2026-04-09 19:01:53 +08:00
README.md added ruflo 2026-04-09 19:01:53 +08:00

Requesty.ai Integration - Planning Documentation

Overview

This directory contains comprehensive planning documentation for integrating Requesty.ai as a new provider in the agentic-flow project.

Status: Planning Complete Implementation Status: Not Started Estimated Effort: 13 hours Risk Level: LOW

What is Requesty.ai?

Requesty.ai is a unified AI gateway providing:

  • Access to 300+ AI models from OpenAI, Anthropic, Google, Meta, DeepSeek, and more
  • OpenAI-compatible API (drop-in replacement)
  • 80% cost savings vs direct Anthropic API
  • Built-in analytics, caching, and auto-routing
  • Enterprise features (zero downtime, failover, load balancing)

Documentation Structure

Read the documents in this order:

1. 00-overview.md - Start Here

Read this first!

  • Executive summary
  • Integration goals
  • Key differentiators vs OpenRouter
  • Strategic benefits
  • Success criteria
  • Risk assessment

Time to read: 5 minutes

2. 01-api-research.md - Technical Details

For developers implementing the integration

  • Complete API specification
  • Authentication methods
  • Request/response schemas
  • Tool calling format
  • Model naming conventions
  • Rate limits and pricing
  • Error handling
  • Comparison with OpenRouter and Anthropic

Time to read: 15 minutes

3. 02-architecture.md - System Design

For architects and lead developers

  • High-level architecture diagrams
  • Component breakdown
  • Data flow diagrams
  • File structure
  • Configuration management
  • Error handling strategy
  • Performance considerations
  • Security architecture

Time to read: 20 minutes

4. 03-implementation-phases.md - Action Plan

For developers ready to implement

  • Step-by-step implementation guide
  • 5 phases with clear deliverables
  • Code examples
  • Acceptance criteria
  • Timeline estimates
  • Post-implementation checklist

Time to read: 25 minutes Implementation time: 13 hours

5. 04-testing-strategy.md - Quality Assurance

For QA engineers and testers

  • Unit test specifications
  • Integration test scenarios
  • E2E user workflows
  • Model-specific tests
  • Performance benchmarks
  • Security tests
  • Acceptance criteria

Time to read: 15 minutes Testing time: 3 hours

6. 05-migration-guide.md - User Documentation

For end users

  • Quick start guide (3 steps)
  • Usage examples
  • Model recommendations
  • Configuration options
  • Migration from other providers
  • Troubleshooting
  • FAQ

Time to read: 10 minutes

Key Findings

High Compatibility with OpenRouter

The research revealed that Requesty.ai uses almost identical API format to OpenRouter:

Aspect OpenRouter Requesty Compatibility
API Format OpenAI /chat/completions OpenAI /chat/completions 100%
Tool Calling OpenAI functions OpenAI functions 100%
Streaming SSE (OpenAI) SSE (OpenAI) 100%
Auth Method Bearer token Bearer token 100%
Request Schema OpenAI OpenAI 100%
Response Schema OpenAI OpenAI 100%

Implication: We can clone the OpenRouter proxy with minimal changes (~95% code reuse).

Implementation Approach

Strategy: Clone and adapt the existing OpenRouter proxy

Effort Breakdown:

  • Phase 1: Core Proxy (4 hours) - Clone OpenRouter proxy
  • Phase 2: CLI Integration (2 hours) - Add provider detection
  • Phase 3: Model Support (2 hours) - Add model definitions
  • Phase 4: Testing (3 hours) - Comprehensive validation
  • Phase 5: Documentation (2 hours) - User guides

Total: 13 hours

Major Benefits

  1. 300+ Models (vs OpenRouter's 100+)
  2. Built-in Analytics (OpenRouter lacks this)
  3. Auto-Routing (intelligent model selection)
  4. Caching (reduce API costs further)
  5. 80% Cost Savings (vs direct Anthropic API)

Risks

Technical Risks: LOW

  • API format is well-documented (OpenAI-compatible)
  • Pattern is proven (OpenRouter already works)
  • 95% code reuse minimizes bugs

Business Risks: LOW

  • Multi-provider architecture already supports fallbacks
  • Users can easily switch providers
  • No vendor lock-in

Quick Reference

Files to Create

agentic-flow/
└── src/
    └── proxy/
        └── anthropic-to-requesty.ts   (~750 lines, 95% from OpenRouter)

Files to Modify

agentic-flow/
├── src/
│   ├── cli-proxy.ts                   (+ ~80 lines)
│   ├── agents/claudeAgent.ts          (+ ~15 lines)
│   └── utils/
│       ├── modelCapabilities.ts       (+ ~50 lines)
│       └── modelOptimizer.ts          (+ ~100 lines)
└── README.md                          (+ Requesty section)

Total Code Impact

Metric Count
New files 1
Modified files 4
New lines of code ~1,000
Reused lines ~750 (95% from OpenRouter)
Original code ~250

Success Criteria

Must Have (MVP)

  • Users can use --provider requesty flag
  • Requesty API key via REQUESTY_API_KEY environment variable
  • Chat completions work with at least 10 tested models
  • Native tool calling support (MCP tools work)
  • Streaming responses supported
  • Error handling and logging
  • Model override via --model flag

Should Have (V1)

  • Tool emulation for models without native support
  • Model capability detection for Requesty models
  • Integration with model optimizer (--optimize)
  • Analytics and usage tracking
  • Proxy mode for Claude Code/Cursor
  • Cost estimation and reporting

Implementation Checklist

Use this checklist when implementing:

Phase 1: Core Proxy Planned

  • Clone anthropic-to-openrouter.ts to anthropic-to-requesty.ts
  • Update class name, base URL, API key variable
  • Update logging messages
  • Test compilation

Phase 2: CLI Integration Planned

  • Add shouldUseRequesty() method
  • Add startRequestyProxy() method
  • Integrate into start flow
  • Update runAgent method
  • Test CLI detection

Phase 3: Model Support Planned

  • Add 15+ models to modelCapabilities.ts
  • Update claudeAgent.ts provider detection
  • Add 10+ models to model optimizer
  • Test model detection

Phase 4: Testing Planned

  • Write unit tests (>90% coverage)
  • Run integration tests (5+ models)
  • Test tool calling
  • Test streaming
  • Validate error handling

Phase 5: Documentation Planned

  • Update README.md
  • Create migration guide
  • Update help text
  • Update .env.example

Next Steps

  1. Review Planning Docs - Read 00-overview.md through 05-migration-guide.md
  2. Get Stakeholder Approval - Present plan to team/maintainers
  3. Set Up Test Account - Get Requesty.ai API key for testing
  4. Begin Implementation - Follow 03-implementation-phases.md
  5. Test Thoroughly - Use 04-testing-strategy.md
  6. Ship to Users - Deploy with 05-migration-guide.md

Questions?

If you have questions about the implementation plan:

  1. Check the FAQ in 05-migration-guide.md
  2. Review the specific planning document
  3. Open a GitHub issue with questions
  4. Tag the planning document author

Contributing

If you find gaps in the planning documentation:

  1. Open an issue describing the gap
  2. Submit a PR with improvements
  3. Update this README with new findings

Changelog

  • 2025-01-07 - Initial planning documentation created
  • Research completed on Requesty.ai API
  • All 6 planning documents written
  • Ready for implementation

Credits

Planning Author: Claude Code Project: agentic-flow Based On: OpenRouter integration pattern Documentation Standard: SPARC methodology


Ready to implement? Start with 03-implementation-phases.md

Need user docs? Jump to 05-migration-guide.md

Want technical details? Read 02-architecture.md