260 lines
6.7 KiB
Markdown
Raw Permalink Normal View History

# vapora-a2a
**Agent-to-Agent (A2A) Protocol Server** - Production-ready implementation of the A2A specification for VAPORA.
## Features
-**Full A2A Protocol Compliance** - JSON-RPC 2.0, Agent Card discovery
-**SurrealDB Persistence** - Tasks survive restarts, production-ready storage
-**NATS Async Coordination** - Real-time task completion via message queue
-**Prometheus Metrics** - Full observability with `/metrics` endpoint
-**Type-Safe** - Rust compile-time guarantees for protocol correctness
-**Integration Tests** - 5 comprehensive end-to-end tests
## Architecture
```
┌─────────────────────────────────────────────────┐
│ A2A HTTP Server (Axum) │
│ /.well-known/agent.json | /a2a | /metrics │
└────────────────┬────────────────────────────────┘
┌───────┴────────┐
│ │
┌────▼─────┐ ┌──────▼────────┐
│ Bridge │ │ TaskManager │
│ (NATS) │ │ (SurrealDB) │
└────┬─────┘ └──────┬────────┘
│ │
│ ┌──────▼────────┐
└────────►│ AgentCoord │
└───────────────┘
```
### Components
1. **CoordinatorBridge** - Maps A2A tasks to internal agent coordination
- NATS subscribers for TaskCompleted/TaskFailed events
- DashMap for async result delivery via oneshot channels
- Graceful degradation if NATS unavailable
2. **TaskManager** - Persistent task storage and lifecycle
- SurrealDB integration with Surreal<Client>
- Parameterized queries for security
- Tasks survive server restarts
3. **Server** - HTTP endpoints (Axum)
- `GET /.well-known/agent.json` - Agent discovery
- `POST /a2a` - Task dispatch
- `GET /a2a/tasks/{task_id}` - Status query
- `GET /health` - Health check
- `GET /metrics` - Prometheus metrics
## Quick Start
### Prerequisites
```bash
# Start SurrealDB
docker run -d -p 8000:8000 \
surrealdb/surrealdb:latest \
start --bind 0.0.0.0:8000
# Start NATS (optional, graceful degradation)
docker run -d -p 4222:4222 nats:latest
# Run migration
surrealdb import --conn ws://localhost:8000 \
--user root --pass root \
migrations/007_a2a_tasks_schema.surql
```
### Run Server
```bash
cargo run --bin vapora-a2a -- \
--host 127.0.0.1 \
--port 8003 \
--version 1.0.0
```
Server will start on `http://127.0.0.1:8003`
### Using the Client
```rust
use vapora_a2a_client::{A2aClient, RetryPolicy};
#[tokio::main]
async fn main() {
let client = A2aClient::new("http://localhost:8003");
// Discover agent capabilities
let agent_card = client.discover_agent().await?;
println!("Agent: {} v{}", agent_card.name, agent_card.version);
// Dispatch task
let task_id = client.dispatch_task(
"task-123".to_string(),
"Write hello world function".to_string(),
Some("In Rust".to_string()),
Some("developer".to_string()),
).await?;
// Poll for completion
loop {
let status = client.get_task_status(&task_id).await?;
match status.state.as_str() {
"completed" => {
println!("Result: {:?}", status.result);
break;
}
"failed" => {
eprintln!("Error: {:?}", status.error);
break;
}
_ => tokio::time::sleep(Duration::from_millis(500)).await,
}
}
}
```
## Configuration
### Environment Variables
```bash
# SurrealDB
SURREAL_URL=ws://localhost:8000
SURREAL_USER=root
SURREAL_PASS=root
# NATS (optional)
NATS_URL=nats://localhost:4222
# Server
A2A_HOST=0.0.0.0
A2A_PORT=8003
A2A_VERSION=1.0.0
```
### CLI Arguments
```bash
vapora-a2a \
--host 0.0.0.0 \
--port 8003 \
--version 1.0.0
```
## Metrics
Exposed at `http://localhost:8003/metrics` in Prometheus text format:
- `vapora_a2a_tasks_total{status="waiting|working|completed|failed"}` - Task counts
- `vapora_a2a_task_duration_seconds{status="completed|failed"}` - Task execution time
- `vapora_a2a_nats_messages_total{subject,result}` - NATS message handling
- `vapora_a2a_db_operations_total{operation,result}` - Database operations
- `vapora_a2a_coordinator_assignments_total{skill,result}` - Coordinator assignments
## Testing
### Unit Tests
```bash
cargo test -p vapora-a2a --lib
```
### Integration Tests
Require SurrealDB + NATS running:
```bash
# Start dependencies
docker compose up -d surrealdb nats
# Run tests
cargo test -p vapora-a2a --test integration_test -- --ignored
# Tests:
# 1. Task persistence after restart
# 2. NATS completion updates DB
# 3. Task state transitions
# 4. Task failure handling
# 5. End-to-end dispatch with timeout
```
## Protocol Compliance
Implements [A2A Protocol Specification](https://a2a-spec.dev):
- ✅ Agent Card (`/.well-known/agent.json`)
- ✅ Task dispatch (`POST /a2a`)
- ✅ Status query (`GET /a2a/tasks/{id}`)
- ✅ JSON-RPC 2.0 envelope
- ✅ Task lifecycle (waiting → working → completed|failed)
- ✅ Artifact support
- ✅ Error handling
## Production Deployment
See [ADR-0001](../../docs/architecture/adr/0001-a2a-protocol-implementation.md) and [ADR-0002](../../docs/architecture/adr/0002-kubernetes-deployment-strategy.md).
### Kubernetes
```bash
kubectl apply -k kubernetes/overlays/prod/
```
### Docker
```bash
docker build -t vapora-a2a:latest -f Dockerfile .
docker run -p 8003:8003 \
-e SURREAL_URL=ws://surrealdb:8000 \
-e NATS_URL=nats://nats:4222 \
vapora-a2a:latest
```
## Troubleshooting
### Tasks not persisting
Check SurrealDB connection:
```bash
# Verify connection
curl http://localhost:8000/health
# Check migration applied
surrealdb sql --conn ws://localhost:8000 \
--user root --pass root --ns test --db main \
"SELECT * FROM a2a_tasks LIMIT 1;"
```
### Tasks not completing
Check NATS connection:
```bash
# Subscribe to completion events
nats sub "vapora.tasks.completed"
# Server will log warnings if NATS unavailable
```
### Metrics not showing
```bash
# Check metrics endpoint
curl http://localhost:8003/metrics | grep vapora_a2a
```
## Related Crates
- **vapora-a2a-client** - Client library for calling A2A servers
- **vapora-agents** - Agent coordinator and registry
- **vapora-backend** - Main VAPORA REST API
## License
MIT OR Apache-2.0