Files
gemini-nexus/.planning/proxy-improvements/03-advanced-features.md
Gemini Nexus Bot a905877c76 feat: bump version to 0.14.0
Includes all changes from this release cycle:
- Top header navigation (sidebar removed)
- Accounts screen decomposition and visual redesign
- Dev/prod environment isolation
- Usage Analytics with accurate averages, cost reporting, OpenRouter pricing
- Monitor moved to Usage with per-request token/cost tracking
- Clickable traffic logs with expandable JSON body viewer
- OpenCode IDE config in Quick Setup
- API Proxy reorganization (flattened tabs)
- CI/CD pipeline fixes (release config, Node version, pnpm alignment)
- Repository cleanup (~80MB of binaries removed)
2026-05-02 00:32:21 -03:00

149 lines
4.9 KiB
Markdown

# Proxy Page — Advanced Features Hidden in Backend
> Generated from oracle agent analysis of `src/server/modules/proxy/`
> Date: 2026-04-22
## Discovery
The backend (`ProxyService`, `ProxyController`, `ProxyModule`) implements many sophisticated features that are **completely invisible** in the current UI. The proxy page only shows: port, protocol, API key, and basic start/stop.
## Hidden Backend Features
### 1. Advanced Routing & Scheduling Controls
**Backend location**: `proxy.service.ts`, `proxy-scheduler.service.ts`
**What's implemented**:
- `scheduler.maxRequestsPerAccount`: Limit requests per account before switching
- `scheduler.maxWaitTime`: Max time to wait for an available account
- `scheduler.retryAttempts`: Number of retries per request
- `scheduler.backoffStrategy`: `exponential` | `linear` | `fixed`
- `scheduler.priorityAccounts`: Array of account IDs to prioritize
**UI exposure**: None. These are hardcoded or set via config files only.
### 2. Live Traffic Inspector
**Backend location**: `traffic-logger.service.ts`
**What's implemented**:
- `TrafficLogger` captures every request/response with metadata
- Fields: timestamp, requestId, accountId, model, promptTokens, completionTokens, latency, status, error
- Log rotation and cleanup policies
- Export to JSON/CSV
**UI exposure**: None. Logs are only written to files.
### 3. Circuit Breaker Dashboard
**Backend location**: `proxy-circuit-breaker.service.ts`
**What's implemented**:
- Per-account circuit breaker state: `CLOSED`, `OPEN`, `HALF_OPEN`
- Failure threshold and recovery timeout
- When open: requests fail fast without hitting the API
- Metrics: failure rate, last failure time, consecutive successes
**UI exposure**: None. Circuit breaker state is internal only.
### 4. Parity Testing Sandbox
**Backend location**: `proxy-parity.service.ts`
**What's implemented**:
- Send identical requests to multiple models simultaneously
- Compare responses for consistency
- Detect model drift or API changes
- Configurable test suites
**UI exposure**: None. Only accessible via manual API calls or scripts.
### 5. Custom Model Mapping Table
**Backend location**: `proxy-model-mapper.service.ts`
**What's implemented**:
- Dynamic model resolution: `claude-opus-4` → `gemini-3-flash`
- User-defined mappings stored in database
- Regex-based model matching
- Fallback chain: if mapped model unavailable, try next
**UI exposure**: A static, read-only list is shown. No editing capability.
### 6. Upstream Proxy Configuration
**Backend location**: `proxy-upstream.service.ts`
**What's implemented**:
- `upstreamProxy.url`: Route all traffic through another proxy
- `upstreamProxy.auth`: Basic auth or bearer token for upstream
- `upstreamProxy.bypassList`: Domains to bypass
**UI exposure**: None. Set via environment variables only.
### 7. IDE Integration Quick-Setup
**Backend location**: `proxy-config.service.ts` — `generateIdeConfig()`
**What's implemented**:
- Generates ready-to-paste configs for:
- Cursor (`settings.json`)
- VS Code + Cline/Roo Code
- OpenCode (`.env`)
- Claude Code (`claude_desktop_config.json`)
- Generic OpenAI client
- Includes local IP detection, port, and API key
**UI exposure**: Code examples are generic curl snippets. No IDE-specific configs.
### 8. Health & Performance Metrics
**Backend location**: `proxy-metrics.service.ts`
**What's implemented**:
- Request latency histogram (p50, p95, p99)
- Requests per minute counter
- Error rate by status code
- Active connection gauge
- Token throughput (tokens/minute)
- Account utilization percentage
**UI exposure**: None. Metrics are only exposed via a `/metrics` endpoint for Prometheus.
### 9. Request Replay & Debug Tool
**Backend location**: `proxy-replay.service.ts`
**What's implemented**:
- Store recent requests (last 100 by default)
- Replay any request with one click
- Modify headers/body before replay
- Compare original vs. replayed response
- Export request as curl command
**UI exposure**: None.
### 10. Global Timeout & Max Wait Controls
**Backend location**: `proxy-config.service.ts`
**What's implemented**:
- `requestTimeout`: Global timeout for all upstream requests
- `connectTimeout`: TCP connection timeout
- `maxWaitTime`: Max time to wait for an available account
- `idleTimeout`: Close idle connections after N seconds
**UI exposure**: None. Hardcoded defaults.
---
## Recommendation
Create a new "Advanced" tab on the proxy page that exposes these features progressively:
1. **Phase 1**: Read-only views (circuit breaker status, health metrics, traffic logs)
2. **Phase 2**: Interactive controls (model mapping editor, scheduler settings)
3. **Phase 3**: Power-user tools (replay, parity testing, upstream proxy config)
This would transform the proxy page from a simple on/off switch into a comprehensive API gateway dashboard.