vrr/ruvector
2
0
Fork 0
mirror of https://github.com/ruvnet/RuVector.git synced 2026-05-25 15:03:46 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
ruvector/docs/research/claude-code-rvsource/10-models-and-api.md
rUv 930fca916f feat(sse): decouple SSE to mcp.pi.ruv.io proxy + Claude Code source research 
SSE Proxy Decoupling (ADR-130):
- Fix ruvbrain-sse proxy: proper MCP handshake, session creation, drain polling
- Fix internal queue endpoints: session_create keeps receiver, drain returns buffered messages
- Add response_queues to AppState for SSE proxy communication
- Skip sparsifier for >5M edge graphs (was crashing on 16M edges)
- Add SSE_DISABLED/MAX_SSE env vars for configurable connection limits
- Route SSE to dedicated mcp.pi.ruv.io subdomain (Cloudflare CNAME)
- Serve SSE at root / path on proxy (no /sse needed)
- Update all references from pi.ruv.io/sse to mcp.pi.ruv.io
- Fix Dockerfile consciousness crate build (feature/version mismatches)

Claude Code CLI Source Research (ADR-133):
- 19 research documents analyzing Claude Code internals (3000+ lines)
- Decompiler script + RVF corpus builder for all major versions
- Binary RVF containers for v0.2, v1.0, v2.0, v2.1 (300-2068 vectors each)
- Call graphs, class hierarchies, state machines from minified source

Integration Strategy (ADR-134):
- 6-tier integration plan: WASM MCP, agents, hooks, cache, SDK, plugin
- Integration guide with architecture diagrams and performance targets

Co-Authored-By: claude-flow <ruv@ruv.net>
2026-04-02 23:39:56 +00:00

5.7 KiB
Raw Permalink Blame History

Claude Code CLI: Models and API Integration

Supported Models

Model References Found in Source

Model ID Family Notes
claude-opus-4-6 Opus Latest (current default for complex)
claude-opus-4-5 Opus
claude-opus-4-5-20251101 Opus Dated release
claude-opus-4-1 Opus
claude-opus-4-1-20250805 Opus Dated release
claude-opus-4-0 Opus
claude-opus-4 Opus Alias
claude-opus-4-20250514 Opus Dated release
claude-4-opus-20250514 Opus Legacy naming
claude-sonnet-4-6 Sonnet Latest Sonnet
claude-sonnet-4-5 Sonnet
claude-sonnet-4-5-20250929 Sonnet Dated release
claude-sonnet-4 Sonnet Alias
claude-sonnet-4-20250514 Sonnet Dated release
claude-sonnet-3-7 Sonnet Legacy
claude-3-7-sonnet-20250219 Sonnet Legacy naming
claude-3-5-sonnet-20241022 Sonnet Legacy
claude-3-sonnet-20240229 Sonnet Legacy
claude-haiku-4-5 Haiku
claude-haiku-4-5-20251001 Haiku Dated release
claude-haiku-4 Haiku Alias
claude-haiku-3-5 Haiku Legacy
claude-3-5-haiku-20241022 Haiku Legacy naming
claude-instant-1.1 Instant Legacy
claude-instant-1.2 Instant Legacy
claude-code-20250219 Code Specialized code model
claude-3-opus-20240229 Opus Legacy

Model Selection

Mechanism Priority Description
--model CLI flag Highest Runtime override
ANTHROPIC_MODEL env var High Environment override
model in settings Medium Persistent config
availableModels allowlist - Restricts options
mainLoopModel - Internal selection
Built-in default Lowest Fallback

Model Aliases

Users can specify short aliases: "opus", "sonnet", "haiku". The availableModels allowlist accepts these aliases, which map to the latest model in each family.

Model Overrides

modelOverrides maps Anthropic model IDs to provider-specific IDs:

{
  "modelOverrides": {
    "claude-sonnet-4-6": "us.anthropic.claude-sonnet-4-6-v1:0"
  }
}

API Integration

Anthropic Direct API

  • Endpoint: ANTHROPIC_BASE_URL (default: https://api.anthropic.com)
  • Authentication: ANTHROPIC_API_KEY or OAuth token
  • Unix socket: ANTHROPIC_UNIX_SOCKET for local proxying

API Endpoints Used

Endpoint Purpose
/v1/messages Main conversation API (streaming)
/v1/messages/count_tokens Token counting
/v1/messages/batches Batch processing
/v1/models Model listing
/v1/complete Legacy completion
/v1/token Token validation
/v1/files File management
/v1/code/upstreamproxy/ws WebSocket proxy
/v2/session_ingress/shttp/mcp/ MCP session ingress

Provider Backends

Provider Client Auth
Anthropic Direct Anthropic (SDK) API key / OAuth
AWS Bedrock BedrockClient / BedrockRuntimeClient AWS IAM
Google Vertex AI Native HTTP GCP credentials
Azure Foundry Native HTTP Azure credentials
Anthropic AWS AnthropicAws Hybrid auth

Prompt Caching

Anthropic's prompt caching reduces repeat token costs:

  • cache_control: { type: "ephemeral" } -- Standard cache
  • 1-hour cache on Bedrock (ENABLE_PROMPT_CACHING_1H_BEDROCK)
  • Per-model disable: DISABLE_PROMPT_CACHING_HAIKU/SONNET/OPUS
  • Cache sharing: promptCacheSharingEnabled
  • Token tracking: promptCacheReadTokens, promptCacheWriteTokens

Retry and Fallback

  • CLAUDE_CODE_MAX_RETRIES -- Max API retry count
  • --fallback-model -- Fallback model for overload (print mode only)
  • FALLBACK_FOR_ALL_PRIMARY_MODELS -- Universal fallback
  • CLAUDE_CODE_SKIP_FAST_MODE_NETWORK_ERRORS -- Skip on network errors
  • CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK -- Disable non-streaming fallback

Request Customization

  • ANTHROPIC_BETAS / --betas -- Beta feature headers
  • ANTHROPIC_CUSTOM_HEADERS -- Custom request headers
  • CLAUDE_CODE_EXTRA_BODY -- Extra request body fields
  • CLAUDE_CODE_EXTRA_METADATA -- Extra metadata
  • API_TIMEOUT_MS -- Request timeout
  • CLAUDE_CODE_ATTRIBUTION_HEADER -- Attribution header
  • CLAUDE_CODE_STALL_TIMEOUT_MS_FOR_TESTING -- Stall detection

Structured Output

  • --json-schema -- Enforce JSON schema on output
  • MAX_STRUCTURED_OUTPUT_RETRIES -- Retry on schema validation failure

Token Management

  • MAX_THINKING_TOKENS -- Cap thinking tokens
  • CLAUDE_CODE_MAX_OUTPUT_TOKENS -- Cap output tokens
  • CLAUDE_CODE_FILE_READ_MAX_OUTPUT_TOKENS -- File read budget
  • CLAUDE_CODE_BLOCKING_LIMIT_OVERRIDE -- Rate limit override
  • Token counting via /v1/messages/count_tokens

Effort Level

  • --effort CLI flag: low, medium, high, max
  • CLAUDE_CODE_EFFORT_LEVEL env var
  • effortLevel in settings
  • CLAUDE_CODE_ALWAYS_ENABLE_EFFORT -- Always apply effort level
  • /effort slash command -- Interactive change

Thinking/Reasoning

  • alwaysThinkingEnabled -- Enable extended thinking
  • CLAUDE_CODE_DISABLE_THINKING -- Disable thinking
  • CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING -- Disable adaptive mode
  • DISABLE_INTERLEAVED_THINKING -- Disable interleaved thinking
  • MAX_THINKING_TOKENS -- Token budget for thinking
  • CLAUDE_CODE_DISABLE_FAST_MODE -- Disable fast mode shortcut

Fast Mode

Fast mode uses smaller/faster models for simple operations:

  • fastMode setting (boolean)
  • fastModePerSessionOptIn -- Opt-in per session
  • /fast slash command -- Toggle
  • ANTHROPIC_SMALL_FAST_MODEL -- Custom small model
  • ANTHROPIC_SMALL_FAST_MODEL_AWS_REGION -- Region for small model