stemedb/ai-lookup/services/meter.md
jordan 1ce4004807 feat: Complete Phase 2 (The Cortex) - query, lens, and API layers
This commit adds the read path (Cortex) to complement the write path (Spine):

## Crates
- stemedb-api: HTTP API with axum + utoipa OpenAPI
  - /v1/assert, /v1/query, /v1/epoch, /v1/skeptic, /v1/trace, /v1/audit
  - Metered endpoints with quota enforcement
  - Ed25519 signature verification
- stemedb-lens: Truth resolution lenses
  - RecencyLens, ConsensusLens, ConfidenceLens
  - VoteAwareConsensusLens (Ballot Box pattern)
  - TrustAwareAuthorityLens (The Hive pattern)
  - SkepticLens (conflict analysis)
  - EpochAwareLens (paradigm-safe queries)
- stemedb-query: Query engine with materialized views

## Storage Extensions
- VoteStore: Vote aggregation with cached counts
- TrustRankStore: Agent reputation with decay
- AuditStore: Query audit trail
- IndexStore: SP/P/S index structures
- SupersessionStore: Epoch supersession chains

## SDKs
- sdk/go/steme: Go HTTP client with Ed25519 signing
- sdk/go/adk: ADK-Go tools for AI agents

## Documentation
- Updated CLAUDE.md, architecture.md, roadmap.md
- New ai-lookup entries for all services
- Use case docs for consumer health intelligence
- Arena roadmap for simulation advancement

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-01 13:22:44 -07:00

132 lines
3.5 KiB
Markdown

# The Meter (Economic Throttling)
## Overview
The Meter implements per-agent per-hour quota enforcement using a Token Bucket algorithm. This prevents runaway agents from exhausting system resources while allowing bursty behavior.
## Cost Model
| Operation | Base Cost | Notes |
|-----------|-----------|-------|
| Assert | 10 tokens | Writing truth is expensive |
| Vote | 1 token | Voting is cheap |
| Query | 5 tokens | + 1 per Lens applied |
| Payload | +1/KB | Prevents spamming massive blobs |
Default quota: **10,000 tokens per agent per hour**.
## Components
### QuotaStore (`stemedb-storage/src/quota_store.rs`)
Storage abstraction for quota tracking.
**Key Types:**
- `QuotaRecord` - Per-agent per-hour usage record
- `CostConfig` - Operation cost configuration
- `QuotaCheckResult` - Result of quota check (allowed, remaining, reset_at)
- `OperationType` - Assert, Vote, or Query
**Storage Layout:**
```
QT:{agent_id}:{hour} -> QuotaRecord (serialized)
QL:{agent_id} -> u64 (custom limit override)
```
**Key Methods:**
```rust
async fn check_and_record(&self, agent_id, operation, payload_bytes, timestamp) -> QuotaCheckResult;
async fn get_quota_status(&self, agent_id, timestamp) -> QuotaCheckResult;
async fn set_quota_limit(&self, agent_id, limit) -> Result<()>;
```
### MeterLayer (`stemedb-api/src/middleware/meter.rs`)
Tower middleware that intercepts requests and enforces quotas.
**Request Flow:**
1. Extract `X-Agent-Id` header (hex-encoded 32-byte public key)
2. Determine operation type from path
3. Calculate cost based on operation + payload size
4. Check quota and reject with 429 if exceeded
5. Record cost and forward request
6. Add quota headers to response
**Headers:**
| Header | Direction | Description |
|--------|-----------|-------------|
| `X-Agent-Id` | Request | Agent's Ed25519 public key (hex, 64 chars) |
| `X-Quota-Remaining` | Response | Remaining tokens in current window |
| `X-Quota-Limit` | Response | Total tokens allowed per hour |
| `X-Quota-Reset` | Response | Unix timestamp when window resets |
### API Endpoints
**GET /v1/meter/quota**
Check quota status for an agent.
Query params:
- `agent_id` - Hex-encoded agent public key
Response:
```json
{
"agent_id": "...",
"remaining": 9500,
"limit": 10000,
"reset_at": 1705317600,
"used": 500,
"window_start": 1705314000
}
```
**POST /v1/meter/quota/limit**
Set custom quota limit (admin operation).
Request:
```json
{
"agent_id": "...",
"limit": 50000
}
```
## Configuration
Environment variable: `STEMEDB_METER_ENABLED`
- `true` (default): Enable economic throttling
- `false` or `0`: Disable (all requests pass through)
## Bypass Paths
These paths bypass metering:
- `/v1/health`
- `/swagger-ui/*`
- `/api-docs/*`
## Usage Example
```bash
# Check quota before operations
curl "http://localhost:3000/v1/meter/quota?agent_id=$(xxd -p -l 32 /dev/urandom)"
# Make request with agent ID
curl -X POST http://localhost:3000/v1/assert \
-H "Content-Type: application/json" \
-H "X-Agent-Id: 0102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f20" \
-d '{"subject": "test", ...}'
# Response headers include quota info
# X-Quota-Remaining: 9989
# X-Quota-Limit: 10000
# X-Quota-Reset: 1705317600
```
## Implementation Notes
- Quota windows are hour-aligned (truncated to the start of the hour)
- Quotas reset automatically at the start of each hour
- On storage errors, requests are allowed (fail open for availability)
- Anonymous requests (no X-Agent-Id) bypass metering
- Token bucket allows burst behavior while enforcing long-term limits