Build AI-Generated Code Your Users Can Rely On
fault is an Open-Source Rust-powered CLI and AI-MCP server that lets engineers and AI agents team up to ship production-ready code.
Plug fault into your favorite AI-Editor
-
| | | |
fault agent tool fault_score_reliability{ "explanation": "The function has no error handling, timeout, retry or fallback logic, making it prone to upstream failures causing unhandled exceptions.", "score": 0.2 }fault agent code-reviewHere’s a small, focused refactor of your `/` handler that will immediately improve resilience, observability and error‐handling without drastically changing your app: 1. **Single, long-lived HTTP client** Re-using an `httpx.Client` gives you connection pooling and lets you centrally configure timeouts. 2. **Per-request timeouts** By setting `timeout=httpx.Timeout(connect=5, read=10)` we fail fast on network issues. 3. **Retries with exponential backoff** The [`backoff`](https://pypi.org/project/backoff/) library wraps `_fetch_todo()` so that transient errors and 5xx status codes automatically retry (up to 5× or ~30 s total). 4. **Error propagation & HTTPExceptions** We catch `RequestError` (network) vs `HTTPStatusError` (non-2xx) and translate them into appropriate 5xx/4xx FastAPI responses. 5. **Logging** Warnings on upstream 4xxs and errors on network failures let you trace the failure mode in your logs. ## Next steps - **Metrics & tracing**: Plug in Prometheus counters or OpenTelemetry spans around `_fetch_todo` to observe retry counts, latencies, and success/fail rates. - **Circuit breaker**: If your upstream is flapping, consider a circuit-breaker pattern (e.g. [`pybreaker`](https://pypi.org/project/pybreaker/)) to avoid overwhelming it. - **Async support**: If you expect high concurrency, convert to `httpx.AsyncClient` and make `index` an `async def`. These minimal changes will immediately make your endpoint far more reliable under network glitches or upstream outages—yet remain straightforward to slot into your existing `app.py`.fault agent tool fault_make_reliability_and_perf_changes--- app.py +++ app.py @@ -1,15 +1,58 @@ import os +import logging import httpx -from fastapi import FastAPI +from fastapi import FastAPI, HTTPException UPSTREAM_URL = os.getenv("UPSTREAM_URL", "https://jsonplaceholder.typicode.com") +REQUEST_TIMEOUT = float(os.getenv("REQUEST_TIMEOUT", 5.0)) +MAX_RETRIES = int(os.getenv("MAX_RETRIES", 3)) +BACKOFF_FACTOR = float(os.getenv("BACKOFF_FACTOR", 0.5)) +RETRY_STATUS_CODES = [502, 503, 504] app = FastAPI() +logger = logging.getLogger("uvicorn.error") +transport = httpx.RetryTransport( + retries=MAX_RETRIES, + backoff_factor=BACKOFF_FACTOR, + status_forcelist=RETRY_STATUS_CODES, + allowed_methods=["GET"], +) +@app.on_event("startup") +async def startup_event(): + ''' + Initialize the HTTPX client with retry and timeout settings for resilience. + ''' + app.state.client = httpx.AsyncClient( + base_url=UPSTREAM_URL, + timeout=httpx.Timeout(REQUEST_TIMEOUT), + transport=transport, + headers={"Host": "jsonplaceholder.typicode.com"}, + ) + +@app.on_event("shutdown") +async def shutdown_event(): + ''' + Close the HTTPX client on application shutdown to release resources. + ''' + await app.state.client.aclose() + @app.get("/") -def index(): - return httpx.get(f"{UPSTREAM_URL}/todos/1", headers={ - "Host": "jsonplaceholder.typicode.com" - }).json() +async def index(): + ''' + Retrieves a todo item from the upstream service. + Implements retry logic, timeouts, and error handling for resilience. + ''' + try: + response = await app.state.client.get("/todos/1") + response.raise_for_status() + except httpx.RequestError as exc: + logger.error(f"Request error: {exc}") + raise HTTPException(status_code=503, detail="Upstream service unavailable") + except httpx.HTTPStatusError as exc: + logger.error(f"Invalid response {exc.response.status_code}: {exc}") + raise HTTPException(status_code=502, detail="Upstream returned an error") + + return response.json()fault agent tool fault_suggest_service_level_objectives_slo[ { "type": "latency", "title": "P95 Latency for /todos/1 API", "objective": 95.0, "explanation": "95th percentile of request latency must stay under threshold to ensure responsiveness", "threshold": 0.3, "unit": "s", "window": "7d", "sli": { "prometheus": "histogram_quantile(0.95, sum(rate(http_request_duration_seconds_bucket{handler=\"/todos/1\"}[5m])) by (le)) * 1000", "gcp/cloudrun": { "displayName": "95% - P95 Latency - Calendar day", "goal": 0.95, "calendarPeriod": "DAY", "serviceLevelIndicator": { "windowsBased": { "windowPeriod": "300s", "goodTotalRatioThreshold": { "basicSliPerformance": { "latency": { "threshold": "0.3s" } }, "threshold": 0.95 } } } } } }, { "type": "availability", "title": "2xx Success Rate for /todos/1 API", "objective": 99.9, "explanation": "Percentage of requests returning HTTP 2xx status codes", "threshold": 99.9, "unit": "", "window": "7d", "sli": { "prometheus": "sum(rate(http_requests_total{handler=\"/todos/1\",status=~\"2..\"}[5m]))/sum(rate(http_requests_total{handler=\"/todos/1\"}[5m]))*100", "gcp/cloudrun": { "displayName": "99.9% - Availability - Calendar day", "goal": 0.999, "calendarPeriod": "DAY", "serviceLevelIndicator": { "windowsBased": { "windowPeriod": "300s", "goodTotalRatioThreshold": { "threshold": 0.999 } } } } } }, { "type": "error", "title": "5xx Error Rate for /todos/1 API", "objective": 1.0, "explanation": "Percentage of requests resulting in HTTP 5xx errors must remain low", "threshold": 1.0, "unit": "%", "window": "7d", "sli": { "prometheus": "sum(rate(http_requests_total{handler=\"/todos/1\",status=~\"5..\"}[5m]))/sum(rate(http_requests_total{handler=\"/todos/1\"}[5m]))*100", "gcp/cloudrun": { "displayName": "1% - 5xx Error Rate - Calendar day", "goal": 0.01, "calendarPeriod": "DAY", "serviceLevelIndicator": { "windowsBased": { "windowPeriod": "300s", "goodTotalRatioThreshold": { "threshold": 0.01 } } } } } } ]fault agent tool fault_run_latency_impact_scenario# Report ## Scenario: Evaluating runtime performance of http://localhost:9090 (items: 1) ### 🎯 `GET` http://localhost:9090 | Passed **Call**: - Method: `GET` - Timeout: 10000ms - Headers: - - Body?: No **Strategy**: load for 10s with 1 clients @ 3 RPS **Faults Applied**: | Type | Timeline | Description | |------|----------|-------------| | latency | 0% `xxxxxxxxxx` 100% | Latency: ➡️🖧, Per Read/Write Op.: false, Distribution: normal, Mean: 300.00 ms, Stddev: 0.00 ms | **Run Overview**: Num. Requests: 29 Num. Errors: 0 (0.0%) Min. Response Time: 334.14 Max Response Time: 462.50 Mean Latency (ms): 339.96 Total Time: 10 seconds and 214 ms | Latency Percentile | Latency (ms) | Num. Requests (% of total) | |------------|--------------|-----------| | p25 | 335.78 | 8 (27.6%) | | p50 | 339.96 | 15 (51.7%) | | p75 | 342.26 | 23 (79.3%) | | p95 | 406.86 | 29 (100.0%) | | p99 | 462.50 | 29 (100.0%) | | SLO | Pass? | Objective | Margin | Num. Requests Over Threshold (% of total) | |-----------|-------|-----------|--------|--------------------------| | 99% @ 350ms | ❌ | 99% < 350ms | Above by 112.5ms | 2 (6.9%) | | 95% @ 200ms | ❌ | 95% < 200ms | Above by 206.9ms | 29 (100.0%) |