traceact 0.3.0__tar.gz → 0.5.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {traceact-0.3.0/traceact.egg-info → traceact-0.5.0}/PKG-INFO +15 -2
- {traceact-0.3.0 → traceact-0.5.0}/README.md +9 -1
- {traceact-0.3.0 → traceact-0.5.0}/USAGE.md +504 -7
- {traceact-0.3.0 → traceact-0.5.0}/pyproject.toml +17 -1
- traceact-0.5.0/tests/test_async_sink.py +212 -0
- traceact-0.5.0/tests/test_http_sink.py +148 -0
- traceact-0.5.0/tests/test_otlp_sink.py +464 -0
- traceact-0.5.0/tests/test_propagation.py +873 -0
- {traceact-0.3.0 → traceact-0.5.0}/tests/test_redaction.py +1 -1
- traceact-0.5.0/tests/test_sqlite_sink.py +172 -0
- traceact-0.5.0/tests/test_tracelog.py +466 -0
- traceact-0.5.0/traceact/__init__.py +112 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/config.py +12 -0
- traceact-0.5.0/traceact/log.py +421 -0
- traceact-0.5.0/traceact/middleware.py +225 -0
- traceact-0.5.0/traceact/propagation.py +267 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/redaction.py +12 -0
- traceact-0.5.0/traceact/sinks.py +1301 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/trace.py +37 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/static/app.js +207 -4
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/static/index.html +24 -4
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/static/styles.css +79 -6
- {traceact-0.3.0 → traceact-0.5.0/traceact.egg-info}/PKG-INFO +15 -2
- {traceact-0.3.0 → traceact-0.5.0}/traceact.egg-info/SOURCES.txt +9 -0
- traceact-0.5.0/traceact.egg-info/requires.txt +9 -0
- traceact-0.3.0/traceact/__init__.py +0 -51
- traceact-0.3.0/traceact/sinks.py +0 -520
- traceact-0.3.0/traceact.egg-info/requires.txt +0 -4
- {traceact-0.3.0 → traceact-0.5.0}/LICENSE +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/MANIFEST.in +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/setup.cfg +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/tests/test_decorators.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/tests/test_doctor.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/tests/test_reader.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/tests/test_sinks.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/budget.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/context.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/decorators.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/helpers.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/ids.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/__init__.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/cli.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/doctor.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/instance.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/reader.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact/viewer/server.py +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact.egg-info/dependency_links.txt +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact.egg-info/entry_points.txt +0 -0
- {traceact-0.3.0 → traceact-0.5.0}/traceact.egg-info/top_level.txt +0 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: traceact
|
|
3
|
-
Version: 0.
|
|
3
|
+
Version: 0.5.0
|
|
4
4
|
Summary: X-ray vision for your code. Lightweight action-level tracing for Python.
|
|
5
5
|
Author: Mohammed Shehu
|
|
6
6
|
License-Expression: MIT
|
|
@@ -26,6 +26,11 @@ License-File: LICENSE
|
|
|
26
26
|
Provides-Extra: dev
|
|
27
27
|
Requires-Dist: pytest>=7.0; extra == "dev"
|
|
28
28
|
Requires-Dist: pytest-asyncio>=0.21; extra == "dev"
|
|
29
|
+
Requires-Dist: pytest-randomly>=3.15; extra == "dev"
|
|
30
|
+
Requires-Dist: flask>=3.0; extra == "dev"
|
|
31
|
+
Requires-Dist: starlette>=0.36; extra == "dev"
|
|
32
|
+
Requires-Dist: django>=4.2; extra == "dev"
|
|
33
|
+
Requires-Dist: requests>=2.31; extra == "dev"
|
|
29
34
|
Dynamic: license-file
|
|
30
35
|
|
|
31
36
|
# TraceAct
|
|
@@ -143,7 +148,15 @@ Checks Python version, that `~/.traceact` is writable, whether a viewer is alrea
|
|
|
143
148
|
| `Step` | A human-readable timeline marker within a trace |
|
|
144
149
|
| `Event` | A structured operation: db, http, file, model, job, etc. |
|
|
145
150
|
| `Touch` | A resource involved in the trace (auto-derived from events) |
|
|
146
|
-
| `Sink` | Where completed traces are written (`JsonlSink`, `ConsoleSink`) |
|
|
151
|
+
| `Sink` | Where completed traces are written (`JsonlSink`, `ConsoleSink`, `AsyncSink`) |
|
|
152
|
+
|
|
153
|
+
### Design principle: observable by choice, never forced blind
|
|
154
|
+
|
|
155
|
+
TraceAct exists to give you X-ray vision for your code. That means nothing TraceAct does itself should take that vision away.
|
|
156
|
+
|
|
157
|
+
Wherever TraceAct might skip, drop, or truncate data — a trace sampled out by `sample_rate`, events truncated by a budget limit, records dropped by `AsyncSink` under backpressure — there is always an observable signal. The `budget_hit` flag marks truncated traces. The `AsyncSink.dropped` counter counts every dropped record. Sampling decisions are made before a trace object is created, so nothing is half-recorded.
|
|
158
|
+
|
|
159
|
+
The design choice is always: **silent by default, observable by choice**. You decide whether to log, alert on, or ignore those signals. TraceAct never makes that decision for you.
|
|
147
160
|
|
|
148
161
|
## Wiring into a web app
|
|
149
162
|
|
|
@@ -113,7 +113,15 @@ Checks Python version, that `~/.traceact` is writable, whether a viewer is alrea
|
|
|
113
113
|
| `Step` | A human-readable timeline marker within a trace |
|
|
114
114
|
| `Event` | A structured operation: db, http, file, model, job, etc. |
|
|
115
115
|
| `Touch` | A resource involved in the trace (auto-derived from events) |
|
|
116
|
-
| `Sink` | Where completed traces are written (`JsonlSink`, `ConsoleSink`) |
|
|
116
|
+
| `Sink` | Where completed traces are written (`JsonlSink`, `ConsoleSink`, `AsyncSink`) |
|
|
117
|
+
|
|
118
|
+
### Design principle: observable by choice, never forced blind
|
|
119
|
+
|
|
120
|
+
TraceAct exists to give you X-ray vision for your code. That means nothing TraceAct does itself should take that vision away.
|
|
121
|
+
|
|
122
|
+
Wherever TraceAct might skip, drop, or truncate data — a trace sampled out by `sample_rate`, events truncated by a budget limit, records dropped by `AsyncSink` under backpressure — there is always an observable signal. The `budget_hit` flag marks truncated traces. The `AsyncSink.dropped` counter counts every dropped record. Sampling decisions are made before a trace object is created, so nothing is half-recorded.
|
|
123
|
+
|
|
124
|
+
The design choice is always: **silent by default, observable by choice**. You decide whether to log, alert on, or ignore those signals. TraceAct never makes that decision for you.
|
|
117
125
|
|
|
118
126
|
## Wiring into a web app
|
|
119
127
|
|
|
@@ -14,9 +14,12 @@ traceact/
|
|
|
14
14
|
context.py — ContextVar for active trace, SKIP sentinel
|
|
15
15
|
redaction.py — SENSITIVE_PATTERNS baseline + REDACTION_PRESETS registry
|
|
16
16
|
sinks.py — JsonlSink (thread-safe, rotation via max_bytes), ConsoleSink,
|
|
17
|
-
AsyncSink (
|
|
17
|
+
AsyncSink (background-thread wrapper; public as of v0.4)
|
|
18
18
|
helpers.py — TraceHelpersMixin (trace.db, trace.http, trace.file, trace.model)
|
|
19
19
|
ids.py — ID generation (trc_, evt_, stp_, corr_ prefixes)
|
|
20
|
+
propagation.py — extract_trace_id, inject_headers, propagate context manager,
|
|
21
|
+
_INCOMING_TRACE_ID ContextVar for cross-service correlation
|
|
22
|
+
middleware.py — TraceActMiddleware (WSGI), TraceActASGIMiddleware (ASGI)
|
|
20
23
|
|
|
21
24
|
viewer/
|
|
22
25
|
cli.py — `traceact view` / `traceact show` / `traceact doctor` CLI entry point
|
|
@@ -52,12 +55,14 @@ tests/ — pytest suite (pip install -e ".[dev]" && pytest)
|
|
|
52
55
|
13. [Background jobs and correlation IDs](#background-jobs-and-correlation-ids)
|
|
53
56
|
14. [Input capture](#input-capture)
|
|
54
57
|
15. [Sinks](#sinks)
|
|
55
|
-
16. [
|
|
56
|
-
17. [
|
|
57
|
-
18. [
|
|
58
|
-
19. [
|
|
59
|
-
20. [
|
|
60
|
-
21. [
|
|
58
|
+
16. [TraceLog](#tracelog)
|
|
59
|
+
17. [Budget configuration](#budget-configuration)
|
|
60
|
+
18. [TraceConfig fields](#traceconfig-fields)
|
|
61
|
+
19. [Test isolation](#test-isolation)
|
|
62
|
+
20. [Trace record schema](#trace-record-schema)
|
|
63
|
+
21. [Viewing traces](#viewing-traces)
|
|
64
|
+
22. [Integrating the viewer into your app](#integrating-the-viewer-into-your-app)
|
|
65
|
+
23. [Distributed propagation](#distributed-propagation)
|
|
61
66
|
|
|
62
67
|
---
|
|
63
68
|
|
|
@@ -688,12 +693,341 @@ configure(sinks=[
|
|
|
688
693
|
])
|
|
689
694
|
```
|
|
690
695
|
|
|
696
|
+
### AsyncSink
|
|
697
|
+
|
|
698
|
+
Wraps any other sink(s) and performs all writes on a background thread, keeping I/O completely off the application's hot path. The traced function drops the record into an in-memory queue and returns immediately; a single worker thread drains the queue at its own pace.
|
|
699
|
+
|
|
700
|
+
```python
|
|
701
|
+
from traceact import AsyncSink, JsonlSink
|
|
702
|
+
|
|
703
|
+
configure(sinks=[
|
|
704
|
+
AsyncSink([JsonlSink("data/traces/traces.jsonl")])
|
|
705
|
+
])
|
|
706
|
+
```
|
|
707
|
+
|
|
708
|
+
**When to use it:** any time the inner sink is slow or remote — an HTTP collector, a database, or a high-latency filesystem. For local files on fast hardware, the overhead of `JsonlSink` alone is rarely worth wrapping.
|
|
709
|
+
|
|
710
|
+
**Backpressure — and why drops are always observable:**
|
|
711
|
+
|
|
712
|
+
If your application produces traces faster than the worker can write them, the queue fills up. TraceAct will never crash or block your app to protect a trace record — but it will never drop records *silently* either. Every record dropped under a backpressure policy is counted in `AsyncSink.dropped`. Check it, log it, or expose it in a health endpoint:
|
|
713
|
+
|
|
714
|
+
```python
|
|
715
|
+
sink = AsyncSink([JsonlSink("traces.jsonl")])
|
|
716
|
+
configure(sinks=[sink])
|
|
717
|
+
|
|
718
|
+
# later, in a health check or periodic log:
|
|
719
|
+
if sink.dropped > 0:
|
|
720
|
+
logger.warning("AsyncSink dropped %d trace records (queue full)", sink.dropped)
|
|
721
|
+
```
|
|
722
|
+
|
|
723
|
+
The whole point of TraceAct is X-ray vision for your code. A silent drop is blindness. The `dropped` counter means you can *choose* to ignore dropped records — but the choice is yours, not the library's.
|
|
724
|
+
|
|
725
|
+
Three policies for when the queue is full:
|
|
726
|
+
|
|
727
|
+
| Policy | Behaviour | Use when |
|
|
728
|
+
|---|---|---|
|
|
729
|
+
| `"drop_newest"` (default) | Discard the incoming record; count it | Older in-flight records are more diagnostically useful |
|
|
730
|
+
| `"drop_oldest"` | Evict the oldest queued record to make room | Recent traces matter more than historical ones |
|
|
731
|
+
| `"block"` | Stall the calling thread until there is space | Zero loss is required and brief latency is acceptable |
|
|
732
|
+
|
|
733
|
+
```python
|
|
734
|
+
AsyncSink([JsonlSink("traces.jsonl")], max_queue=50_000, on_full="drop_oldest")
|
|
735
|
+
```
|
|
736
|
+
|
|
737
|
+
**Graceful shutdown:** `AsyncSink` registers an `atexit` hook on first write. When the process exits normally, the worker flushes all buffered records before stopping — short-lived scripts don't need to call `close()` explicitly, but you can call it yourself at a known shutdown point to flush sooner.
|
|
738
|
+
|
|
739
|
+
**Fork safety:** `os.fork()` doesn't copy background threads into child processes. `AsyncSink` registers a post-fork handler to reset the worker in the child so it starts fresh on the next write.
|
|
740
|
+
|
|
741
|
+
**Wrapping multiple sinks:**
|
|
742
|
+
|
|
743
|
+
```python
|
|
744
|
+
AsyncSink([
|
|
745
|
+
JsonlSink("data/traces/traces.jsonl"),
|
|
746
|
+
ConsoleSink(pretty=False),
|
|
747
|
+
])
|
|
748
|
+
```
|
|
749
|
+
|
|
750
|
+
Both inner sinks receive every record on the background thread. A failing inner sink is caught and skipped so one bad sink can't kill the worker or lose records destined for the others.
|
|
751
|
+
|
|
752
|
+
### SqliteSink
|
|
753
|
+
|
|
754
|
+
Writes finished traces to a local SQLite database using stdlib `sqlite3` — no extra dependencies. Common fields (`action`, `kind`, `status`, `started_at`, `correlation_id`, etc.) are stored as indexed scalar columns for fast filtering; the full trace record is also stored as JSON in a `record` column so no detail is ever lost. The schema is created automatically on first write.
|
|
755
|
+
|
|
756
|
+
```python
|
|
757
|
+
from traceact import SqliteSink, configure
|
|
758
|
+
|
|
759
|
+
configure(sinks=[SqliteSink("data/traces.db")])
|
|
760
|
+
```
|
|
761
|
+
|
|
762
|
+
**Custom table name:**
|
|
763
|
+
|
|
764
|
+
```python
|
|
765
|
+
SqliteSink("data/traces.db", table="my_traces")
|
|
766
|
+
```
|
|
767
|
+
|
|
768
|
+
**Querying traces directly from the database:**
|
|
769
|
+
|
|
770
|
+
```python
|
|
771
|
+
import sqlite3, json
|
|
772
|
+
|
|
773
|
+
conn = sqlite3.connect("data/traces.db")
|
|
774
|
+
|
|
775
|
+
# All failures in the last hour:
|
|
776
|
+
rows = conn.execute("""
|
|
777
|
+
SELECT action, duration_ms, record
|
|
778
|
+
FROM traces
|
|
779
|
+
WHERE status = 'failed'
|
|
780
|
+
AND started_at > datetime('now', '-1 hour')
|
|
781
|
+
ORDER BY started_at DESC
|
|
782
|
+
""").fetchall()
|
|
783
|
+
|
|
784
|
+
for action, ms, raw in rows:
|
|
785
|
+
record = json.loads(raw)
|
|
786
|
+
print(f"{action} {ms}ms errors={record.get('errors')}")
|
|
787
|
+
```
|
|
788
|
+
|
|
789
|
+
**Concurrent writes:** SQLite is opened in WAL mode so reads and writes can proceed concurrently. For high-concurrency workloads, wrap in `AsyncSink` so write latency stays off the application's hot path:
|
|
790
|
+
|
|
791
|
+
```python
|
|
792
|
+
from traceact import AsyncSink, SqliteSink, configure
|
|
793
|
+
|
|
794
|
+
configure(sinks=[AsyncSink([SqliteSink("data/traces.db")])])
|
|
795
|
+
```
|
|
796
|
+
|
|
797
|
+
**Write errors** are printed to stderr and never propagated to the caller — a database hiccup doesn't interrupt the traced function.
|
|
798
|
+
|
|
799
|
+
### HttpSink
|
|
800
|
+
|
|
801
|
+
POSTs each finished trace as a JSON body to an HTTP or HTTPS endpoint. Uses stdlib `urllib` only — zero extra dependencies.
|
|
802
|
+
|
|
803
|
+
```python
|
|
804
|
+
from traceact import HttpSink, AsyncSink, configure
|
|
805
|
+
|
|
806
|
+
configure(sinks=[
|
|
807
|
+
AsyncSink([HttpSink("https://collector.example.com/traces")])
|
|
808
|
+
])
|
|
809
|
+
```
|
|
810
|
+
|
|
811
|
+
**Always wrap in `AsyncSink` for production use.** Each write makes a synchronous HTTP request; without `AsyncSink` that latency hits every traced function call on the return path.
|
|
812
|
+
|
|
813
|
+
**Custom headers** (API keys, auth tokens):
|
|
814
|
+
|
|
815
|
+
```python
|
|
816
|
+
HttpSink(
|
|
817
|
+
"https://collector.example.com/traces",
|
|
818
|
+
headers={"Authorization": "Bearer <your-token>"},
|
|
819
|
+
)
|
|
820
|
+
```
|
|
821
|
+
|
|
822
|
+
**Custom timeout** (default: 5 seconds):
|
|
823
|
+
|
|
824
|
+
```python
|
|
825
|
+
HttpSink("https://collector.example.com/traces", timeout=2.0)
|
|
826
|
+
```
|
|
827
|
+
|
|
828
|
+
**Observable failures:** network errors, timeouts, and non-2xx responses are counted in `HttpSink.failed` — never raised, never silently swallowed. Check it in a health endpoint or periodic log:
|
|
829
|
+
|
|
830
|
+
```python
|
|
831
|
+
sink = HttpSink("https://collector.example.com/traces")
|
|
832
|
+
configure(sinks=[AsyncSink([sink])])
|
|
833
|
+
|
|
834
|
+
# in a health check or periodic log:
|
|
835
|
+
if sink.failed > 0:
|
|
836
|
+
logger.warning("HttpSink: %d trace deliveries failed", sink.failed)
|
|
837
|
+
```
|
|
838
|
+
|
|
839
|
+
### OtlpSink
|
|
840
|
+
|
|
841
|
+
Exports finished traces to any OTLP-compatible collector — Jaeger, Grafana Tempo, Honeycomb, Datadog agent, the OpenTelemetry Collector, and others. Uses OTLP/HTTP+JSON over stdlib `urllib`. Zero extra dependencies; no `opentelemetry-sdk` required.
|
|
842
|
+
|
|
843
|
+
```python
|
|
844
|
+
from traceact import OtlpSink, AsyncSink, configure
|
|
845
|
+
|
|
846
|
+
configure(sinks=[
|
|
847
|
+
AsyncSink([OtlpSink("http://localhost:4318")])
|
|
848
|
+
])
|
|
849
|
+
```
|
|
850
|
+
|
|
851
|
+
Point `endpoint` at your collector's OTLP HTTP receiver base URL (the standard port is 4318). TraceAct appends `/v1/traces` automatically.
|
|
852
|
+
|
|
853
|
+
**Always wrap in `AsyncSink` for production use** — each write makes a synchronous HTTP request.
|
|
854
|
+
|
|
855
|
+
**SaaS collectors (Honeycomb, Datadog, etc.):**
|
|
856
|
+
|
|
857
|
+
```python
|
|
858
|
+
# Honeycomb
|
|
859
|
+
OtlpSink(
|
|
860
|
+
"https://api.honeycomb.io",
|
|
861
|
+
headers={"x-honeycomb-team": "<your-api-key>"},
|
|
862
|
+
)
|
|
863
|
+
|
|
864
|
+
# Datadog (OTLP agent receiver, default port 4318)
|
|
865
|
+
OtlpSink("http://localhost:4318")
|
|
866
|
+
|
|
867
|
+
# Grafana Cloud
|
|
868
|
+
OtlpSink(
|
|
869
|
+
"https://<your-instance>.grafana.net/otlp",
|
|
870
|
+
headers={"Authorization": "Basic <base64-encoded-credentials>"},
|
|
871
|
+
)
|
|
872
|
+
```
|
|
873
|
+
|
|
874
|
+
**Service name and resource attributes:**
|
|
875
|
+
|
|
876
|
+
```python
|
|
877
|
+
OtlpSink(
|
|
878
|
+
"http://localhost:4318",
|
|
879
|
+
resource_attributes={
|
|
880
|
+
"service.name": "orders-api",
|
|
881
|
+
"deployment.env": "production",
|
|
882
|
+
"service.version": "2.1.0",
|
|
883
|
+
},
|
|
884
|
+
)
|
|
885
|
+
```
|
|
886
|
+
|
|
887
|
+
These appear as resource-level attributes on every span exported by this sink. `service.name` defaults to `"traceact"` if you don't set it.
|
|
888
|
+
|
|
889
|
+
**How TraceAct records map to OTel spans:**
|
|
890
|
+
|
|
891
|
+
| TraceAct field | OTel span field |
|
|
892
|
+
|---|---|
|
|
893
|
+
| `action` | Span name |
|
|
894
|
+
| `kind` (`db`, `http`, `cache`, `model`, `auth`, `payment`) | SpanKind CLIENT |
|
|
895
|
+
| `kind` (`queue`, `email`, `export`) | SpanKind PRODUCER |
|
|
896
|
+
| `kind` (`job`) | SpanKind CONSUMER |
|
|
897
|
+
| Everything else | SpanKind INTERNAL |
|
|
898
|
+
| `started_at` / `ended_at` | `startTimeUnixNano` / `endTimeUnixNano` |
|
|
899
|
+
| `status = "completed"` | StatusCode OK |
|
|
900
|
+
| `status = "failed"` | StatusCode ERROR |
|
|
901
|
+
| `parent_trace_id` | `parentSpanId` |
|
|
902
|
+
| `steps` | Span events (`name="step"`) |
|
|
903
|
+
| `errors` | Span events (`name="exception"`) |
|
|
904
|
+
| `inputs.*` | Span attributes `traceact.input.*` |
|
|
905
|
+
| `outputs.*` | Span attributes `traceact.output.*` |
|
|
906
|
+
| `touches` | Span attributes `traceact.touch.N.kind/target` |
|
|
907
|
+
| `trace_id`, `correlation_id`, `actor`, etc. | Span attributes `traceact.*` |
|
|
908
|
+
| Any unlisted scalar field | Span attribute `traceact.<field>` |
|
|
909
|
+
|
|
910
|
+
TraceAct IDs (`trc_...`) are hashed with MD5 to produce the 128-bit trace ID and 64-bit span ID that OTel requires. The original TraceAct IDs are always preserved as `traceact.trace_id` (and `traceact.root_trace_id`, `traceact.correlation_id`) span attributes so you can cross-reference them.
|
|
911
|
+
|
|
912
|
+
**Observable failures:** network errors, timeouts, and non-2xx responses from the collector are counted in `OtlpSink.failed`:
|
|
913
|
+
|
|
914
|
+
```python
|
|
915
|
+
sink = OtlpSink("http://localhost:4318")
|
|
916
|
+
configure(sinks=[AsyncSink([sink])])
|
|
917
|
+
|
|
918
|
+
if sink.failed > 0:
|
|
919
|
+
logger.warning("OtlpSink: %d trace deliveries failed", sink.failed)
|
|
920
|
+
```
|
|
921
|
+
|
|
691
922
|
### Fallback
|
|
692
923
|
|
|
693
924
|
If no sinks are configured when a trace finishes, TraceAct falls back to `ConsoleSink()` so traces aren't silently dropped.
|
|
694
925
|
|
|
695
926
|
---
|
|
696
927
|
|
|
928
|
+
## TraceLog
|
|
929
|
+
|
|
930
|
+
`TraceLog` is the programmatic query interface for TraceAct JSONL files. Use it when code — an AI agent, a test suite, or a background script — needs to read trace data without opening a browser.
|
|
931
|
+
|
|
932
|
+
```python
|
|
933
|
+
from traceact import TraceLog
|
|
934
|
+
|
|
935
|
+
log = TraceLog("data/traces/traces.jsonl") # file or folder
|
|
936
|
+
```
|
|
937
|
+
|
|
938
|
+
A folder source behaves the same as in the viewer: all `.jsonl` files inside it are merged on every read.
|
|
939
|
+
|
|
940
|
+
### Filtering
|
|
941
|
+
|
|
942
|
+
`filter()` returns a **new** `TraceLog` — the original is never mutated.
|
|
943
|
+
|
|
944
|
+
```python
|
|
945
|
+
failures = log.filter(status="failed")
|
|
946
|
+
db_traces = log.filter(kind="db")
|
|
947
|
+
|
|
948
|
+
# AND logic: both conditions must hold
|
|
949
|
+
recent_db_failures = log.filter(kind="db", status="failed")
|
|
950
|
+
|
|
951
|
+
# Chained calls are equivalent
|
|
952
|
+
same_thing = log.filter(kind="db").filter(status="failed")
|
|
953
|
+
```
|
|
954
|
+
|
|
955
|
+
**Supported operators:**
|
|
956
|
+
|
|
957
|
+
| Syntax | Behaviour |
|
|
958
|
+
|---|---|
|
|
959
|
+
| `field=value` | Exact equality (case-sensitive) |
|
|
960
|
+
| `field__contains=value` | Case-insensitive substring |
|
|
961
|
+
| `field__startswith=value` | Case-insensitive prefix |
|
|
962
|
+
| `field__endswith=value` | Case-insensitive suffix |
|
|
963
|
+
| `field__re=pattern` | `re.search` — partial regex match |
|
|
964
|
+
|
|
965
|
+
```python
|
|
966
|
+
log.filter(action__contains="order") # any action with "order"
|
|
967
|
+
log.filter(action__startswith="payment") # actions starting with "payment"
|
|
968
|
+
log.filter(action__re=r"^order\.(create|update)$")
|
|
969
|
+
log.filter(correlation_id="job_abc123") # find one background job's traces
|
|
970
|
+
```
|
|
971
|
+
|
|
972
|
+
### Terminal methods
|
|
973
|
+
|
|
974
|
+
```python
|
|
975
|
+
log.filter(status="failed").all() # List[dict], oldest-first
|
|
976
|
+
log.filter(status="failed").last(10) # 10 most recent
|
|
977
|
+
log.filter(status="failed").first(10) # 10 oldest
|
|
978
|
+
log.filter(status="failed").count() # int
|
|
979
|
+
|
|
980
|
+
log.filter(status="failed").render_table() # pretty-print to stdout
|
|
981
|
+
log.filter(status="failed").render_table(n=25) # cap rows shown
|
|
982
|
+
```
|
|
983
|
+
|
|
984
|
+
Each terminal call re-reads the JSONL file(s) — there is no caching, so you always get the current state of a live source.
|
|
985
|
+
|
|
986
|
+
### Using TraceLog in tests
|
|
987
|
+
|
|
988
|
+
```python
|
|
989
|
+
import pytest
|
|
990
|
+
from traceact import TraceLog, configure, reset_config, JsonlSink
|
|
991
|
+
|
|
992
|
+
def test_checkout_records_payment_touch(tmp_path):
|
|
993
|
+
sink_path = tmp_path / "traces.jsonl"
|
|
994
|
+
configure(sinks=[JsonlSink(str(sink_path))])
|
|
995
|
+
|
|
996
|
+
# run the function under test
|
|
997
|
+
checkout(user_id="u_42", amount=99.00)
|
|
998
|
+
|
|
999
|
+
log = TraceLog(str(sink_path))
|
|
1000
|
+
traces = log.filter(action="checkout").all()
|
|
1001
|
+
|
|
1002
|
+
assert len(traces) == 1
|
|
1003
|
+
touch_kinds = [t["kind"] for t in traces[0].get("touches", [])]
|
|
1004
|
+
assert "payment" in touch_kinds
|
|
1005
|
+
|
|
1006
|
+
reset_config()
|
|
1007
|
+
```
|
|
1008
|
+
|
|
1009
|
+
### TraceLog.view() — shared lens
|
|
1010
|
+
|
|
1011
|
+
`view()` opens the viewer pre-filtered to match the `TraceLog`'s current filters. The viewer shows each active filter as a dismissable badge above the trace list — a human can remove any badge to widen the view, and the search box still works on top.
|
|
1012
|
+
|
|
1013
|
+
```python
|
|
1014
|
+
# Open the viewer showing only failed traces
|
|
1015
|
+
TraceLog("data/traces/traces.jsonl").filter(status="failed").view()
|
|
1016
|
+
|
|
1017
|
+
# Open with multiple filters
|
|
1018
|
+
TraceLog("data/traces/traces.jsonl") \
|
|
1019
|
+
.filter(kind="db", status="failed") \
|
|
1020
|
+
.view()
|
|
1021
|
+
|
|
1022
|
+
# Get the URL without opening a browser (useful in CI or a script)
|
|
1023
|
+
url = log.filter(status="failed").view(open_browser=False)
|
|
1024
|
+
print(f"Open the viewer at: {url}")
|
|
1025
|
+
```
|
|
1026
|
+
|
|
1027
|
+
The viewer is launched (or an existing instance is reused) and pointed at the same source file or folder that the `TraceLog` reads from. The viewer's normal behaviour when opened without `view()` is completely unchanged.
|
|
1028
|
+
|
|
1029
|
+
---
|
|
1030
|
+
|
|
697
1031
|
## Budget configuration
|
|
698
1032
|
|
|
699
1033
|
`TraceBudget` controls how much TraceAct records. When a limit is reached, `budget_hit` is set to `True` on the trace and recording stops. The wrapped function continues running normally.
|
|
@@ -1134,6 +1468,165 @@ Because `launch_or_connect` starts the viewer on `127.0.0.1` by default, the ret
|
|
|
1134
1468
|
|
|
1135
1469
|
---
|
|
1136
1470
|
|
|
1471
|
+
## Distributed propagation
|
|
1472
|
+
|
|
1473
|
+
When a traced action in Service A calls Service B, TraceAct links the two
|
|
1474
|
+
services' traces together with two separate HTTP headers — separate because
|
|
1475
|
+
they answer separate questions:
|
|
1476
|
+
|
|
1477
|
+
| Header | Sets on the receiving trace | Answers |
|
|
1478
|
+
|---|---|---|
|
|
1479
|
+
| `traceact-trace-id` | `upstream_trace_id` | "which trace in another service triggered me?" (causal lineage) |
|
|
1480
|
+
| `traceact-correlation-id` | `correlation_id` | "which wider workflow do I belong to?" (business grouping, passed through untouched) |
|
|
1481
|
+
|
|
1482
|
+
Keeping them separate matters: a trace can have an upstream parent in one
|
|
1483
|
+
service *and* belong to a correlation group that was assigned several hops
|
|
1484
|
+
earlier. Folding one into the other silently discards whichever one loses.
|
|
1485
|
+
|
|
1486
|
+
### Outbound: stamp both headers
|
|
1487
|
+
|
|
1488
|
+
`inject_headers()` reads the active trace and stamps its `trace_id` (always)
|
|
1489
|
+
and its `correlation_id` (if set) onto an outbound headers dict. It never
|
|
1490
|
+
mutates the dict you pass in.
|
|
1491
|
+
|
|
1492
|
+
```python
|
|
1493
|
+
import requests
|
|
1494
|
+
from traceact import inject_headers
|
|
1495
|
+
|
|
1496
|
+
with ActionTrace.start(action="order.submit", correlation_id="corr_wf_9f2a") as trace:
|
|
1497
|
+
headers = inject_headers({"Content-Type": "application/json"})
|
|
1498
|
+
requests.post("https://payments.internal/charge", json=payload, headers=headers)
|
|
1499
|
+
```
|
|
1500
|
+
|
|
1501
|
+
Works the same with `httpx`, `urllib`, or any other HTTP client — `inject_headers`
|
|
1502
|
+
just returns a plain dict.
|
|
1503
|
+
|
|
1504
|
+
### Inbound: extract the headers (manual)
|
|
1505
|
+
|
|
1506
|
+
Use the `propagate` context manager when you want explicit control, or when
|
|
1507
|
+
your framework isn't covered by the automatic middleware. Pass the framework's
|
|
1508
|
+
header object **directly** — `request.headers` works as-is on Flask, Django,
|
|
1509
|
+
FastAPI, and Starlette, and so does a plain dict in any casing:
|
|
1510
|
+
|
|
1511
|
+
```python
|
|
1512
|
+
from traceact import propagate, ActionTrace
|
|
1513
|
+
|
|
1514
|
+
def handle_charge(request):
|
|
1515
|
+
with propagate(request.headers):
|
|
1516
|
+
with ActionTrace.start(action="charge.process") as trace:
|
|
1517
|
+
# trace.upstream_trace_id == Service A's trace_id
|
|
1518
|
+
# trace.correlation_id == whatever correlation Service A had
|
|
1519
|
+
...
|
|
1520
|
+
```
|
|
1521
|
+
|
|
1522
|
+
HTTP header names are case-insensitive and every framework reconstructs them
|
|
1523
|
+
differently (Flask/Werkzeug and Django hand back `Traceact-Trace-Id`,
|
|
1524
|
+
Title-Case; ASGI delivers raw lowercase bytes). `propagate()` and
|
|
1525
|
+
`extract_trace_id()`/`extract_correlation_id()` normalise all of that
|
|
1526
|
+
internally — pass the header object however your framework gives it to you,
|
|
1527
|
+
including `dict(request.headers)` if you already have that.
|
|
1528
|
+
|
|
1529
|
+
### Inbound: automatic via middleware (Flask / Django)
|
|
1530
|
+
|
|
1531
|
+
```python
|
|
1532
|
+
from traceact import TraceActMiddleware
|
|
1533
|
+
|
|
1534
|
+
# Flask
|
|
1535
|
+
app.wsgi_app = TraceActMiddleware(app.wsgi_app)
|
|
1536
|
+
|
|
1537
|
+
# Django (in wsgi.py)
|
|
1538
|
+
from django.core.wsgi import get_wsgi_application
|
|
1539
|
+
application = TraceActMiddleware(get_wsgi_application())
|
|
1540
|
+
```
|
|
1541
|
+
|
|
1542
|
+
Reads both headers from the WSGI environ and applies them to every trace
|
|
1543
|
+
started during that request — including traces started while a streamed
|
|
1544
|
+
response body is generated (the context is held until the response is fully
|
|
1545
|
+
closed, not just until the view function returns). If neither header is
|
|
1546
|
+
present, the middleware is fully transparent.
|
|
1547
|
+
|
|
1548
|
+
### Inbound: automatic via middleware (FastAPI / Starlette)
|
|
1549
|
+
|
|
1550
|
+
```python
|
|
1551
|
+
from traceact import TraceActASGIMiddleware
|
|
1552
|
+
from fastapi import FastAPI
|
|
1553
|
+
|
|
1554
|
+
app = FastAPI()
|
|
1555
|
+
app.add_middleware(TraceActASGIMiddleware)
|
|
1556
|
+
```
|
|
1557
|
+
|
|
1558
|
+
Or wrap manually:
|
|
1559
|
+
|
|
1560
|
+
```python
|
|
1561
|
+
app = TraceActASGIMiddleware(app)
|
|
1562
|
+
```
|
|
1563
|
+
|
|
1564
|
+
### How the chain looks in the viewer and TraceLog
|
|
1565
|
+
|
|
1566
|
+
Both services write their own trace records. Query by whichever link answers
|
|
1567
|
+
your question — `upstream_trace_id` for "what did this specific call trigger",
|
|
1568
|
+
`correlation_id` for "show me the whole workflow":
|
|
1569
|
+
|
|
1570
|
+
```python
|
|
1571
|
+
from traceact import TraceLog
|
|
1572
|
+
|
|
1573
|
+
log = TraceLog("traces.jsonl")
|
|
1574
|
+
log.filter(upstream_trace_id="trc_abc123").all() # traces this call triggered
|
|
1575
|
+
log.filter(correlation_id="corr_wf_9f2a").all() # the entire workflow
|
|
1576
|
+
log.filter(correlation_id="corr_wf_9f2a").view() # same, in the browser viewer
|
|
1577
|
+
```
|
|
1578
|
+
|
|
1579
|
+
The inspector's summary card shows both when present, each in full (not
|
|
1580
|
+
shortened) so they can be copied and searched against another service's logs.
|
|
1581
|
+
|
|
1582
|
+
### `ai_prompts` redaction preset
|
|
1583
|
+
|
|
1584
|
+
For AI pipelines where trace payloads must not store raw prompt text, model
|
|
1585
|
+
responses, or conversation history, enable the `ai_prompts` preset:
|
|
1586
|
+
|
|
1587
|
+
```python
|
|
1588
|
+
configure(
|
|
1589
|
+
config=TraceConfig(redaction_presets=["ai_prompts"]),
|
|
1590
|
+
)
|
|
1591
|
+
```
|
|
1592
|
+
|
|
1593
|
+
Redacted fields include: `raw_prompt`, `prompt_content`, `system_prompt`,
|
|
1594
|
+
`raw_response`, `response_content`, `conversation`, `message_content`,
|
|
1595
|
+
`file_content`, `source_excerpt`, `context_window`, `completion`,
|
|
1596
|
+
`generation`, `output_text`. Safe fields like `model`, `latency_ms`,
|
|
1597
|
+
`prompt_id`, and token counts are unaffected.
|
|
1598
|
+
|
|
1599
|
+
| What gets redacted | What stays |
|
|
1600
|
+
|---|---|
|
|
1601
|
+
| Raw prompts and responses | Model name, version |
|
|
1602
|
+
| System prompts | Token counts (as long as field name has no "token" substring) |
|
|
1603
|
+
| Conversation history | Latency, cost |
|
|
1604
|
+
| File and source excerpts | Trace IDs, correlation IDs |
|
|
1605
|
+
|
|
1606
|
+
Combine with other presets:
|
|
1607
|
+
|
|
1608
|
+
```python
|
|
1609
|
+
TraceConfig(redaction_presets=["ai_prompts", "api_keys"])
|
|
1610
|
+
```
|
|
1611
|
+
|
|
1612
|
+
### Reference: headers
|
|
1613
|
+
|
|
1614
|
+
| Header | Outbound (inject) | Inbound (extract) |
|
|
1615
|
+
|---|---|---|
|
|
1616
|
+
| `traceact-trace-id` | Active `ActionTrace.trace_id` | Sets `upstream_trace_id` |
|
|
1617
|
+
| `traceact-correlation-id` | Active `ActionTrace.correlation_id`, if set | Sets `correlation_id` |
|
|
1618
|
+
|
|
1619
|
+
Accepted header collection types on the inbound side: any object with
|
|
1620
|
+
`.items()` (dict in any casing, Werkzeug/Django/Starlette header objects,
|
|
1621
|
+
`requests.CaseInsensitiveDict`), a list of `(name, value)` pairs, or raw ASGI
|
|
1622
|
+
`[(b"name", b"value")]` byte pairs. Matching is case-insensitive regardless of
|
|
1623
|
+
which form you pass.
|
|
1624
|
+
|
|
1625
|
+
WSGI: the header arrives as `HTTP_TRACEACT_TRACE_ID` in the environ.
|
|
1626
|
+
ASGI: the header arrives as bytes in `scope["headers"]`.
|
|
1627
|
+
|
|
1628
|
+
---
|
|
1629
|
+
|
|
1137
1630
|
## Quick reference
|
|
1138
1631
|
|
|
1139
1632
|
```python
|
|
@@ -1146,5 +1639,9 @@ from traceact import (
|
|
|
1146
1639
|
traced_action, # decorator
|
|
1147
1640
|
JsonlSink, # write traces to a .jsonl file
|
|
1148
1641
|
ConsoleSink, # print traces to stdout
|
|
1642
|
+
propagate, # context manager for inbound propagation
|
|
1643
|
+
inject_headers, # stamp outbound request headers
|
|
1644
|
+
TraceActMiddleware, # WSGI auto-propagation (Flask, Django)
|
|
1645
|
+
TraceActASGIMiddleware, # ASGI auto-propagation (FastAPI, Starlette)
|
|
1149
1646
|
)
|
|
1150
1647
|
```
|
|
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
|
|
|
4
4
|
|
|
5
5
|
[project]
|
|
6
6
|
name = "traceact"
|
|
7
|
-
version = "0.
|
|
7
|
+
version = "0.5.0"
|
|
8
8
|
description = "X-ray vision for your code. Lightweight action-level tracing for Python."
|
|
9
9
|
readme = "README.md"
|
|
10
10
|
requires-python = ">=3.9"
|
|
@@ -35,6 +35,22 @@ dependencies = []
|
|
|
35
35
|
dev = [
|
|
36
36
|
"pytest>=7.0",
|
|
37
37
|
"pytest-asyncio>=0.21",
|
|
38
|
+
# Test-order randomisation. Activates automatically once installed — no
|
|
39
|
+
# flag needed. Kept on deliberately: an order-dependent failure means
|
|
40
|
+
# shared/global state is leaking between tests, which is a real bug.
|
|
41
|
+
# Never pin --randomly-seed in CI to "stabilise" a run; that hides exactly
|
|
42
|
+
# the failure mode this plugin exists to catch.
|
|
43
|
+
"pytest-randomly>=3.15",
|
|
44
|
+
# Real framework objects for propagation tests (extract_trace_id,
|
|
45
|
+
# TraceActMiddleware, TraceActASGIMiddleware). These are test-only —
|
|
46
|
+
# traceact itself has zero runtime dependencies. Testing against a
|
|
47
|
+
# hand-built dict instead of an actual Flask/Django/Starlette request
|
|
48
|
+
# object was exactly the gap that let a header-parsing bug through
|
|
49
|
+
# 43 green tests; these are here so that never happens again.
|
|
50
|
+
"flask>=3.0",
|
|
51
|
+
"starlette>=0.36",
|
|
52
|
+
"django>=4.2",
|
|
53
|
+
"requests>=2.31",
|
|
38
54
|
]
|
|
39
55
|
|
|
40
56
|
[project.urls]
|