Produce a reproducible, sanitized diagnostic bundle for a LangChain / LangGraph
incident — environment snapshot, version manifest, filtered astream_events(v2)
transcript, propagating callback stack, LangSmith trace URL — so a debug
colleague can reproduce the failure without a live terminal. Use when triaging
a production incident, filing a Discord or GitHub bug report, asking for help
on the LangChain forum, or archiving a post-mortem artifact.
Trigger with "langchain debug bundle", "langgraph debug dump",
"langchain diagnostic export", "langsmith trace export", "astream_events dump",
"langchain incident bundle".
Produce a reproducible, sanitized diagnostic bundle for a LangChain / LangGraph
incident — environment snapshot, version manifest, filtered astream_events(v2)
transcript, propagating callback stack, LangSmith trace URL — so a debug
colleague can reproduce the failure without a live terminal. Use when triaging
a production incident, filing a Discord or GitHub bug report, asking for help
on the LangChain forum, or archiving a post-mortem artifact.
Trigger with "langchain debug bundle", "langgraph debug dump",
"langchain diagnostic export", "langsmith trace export", "astream_events dump",
"langchain incident bundle".
An on-call engineer pages you at 2am: the production agent loops, ToolMessage
outputs are empty strings, the user sees "I could not find the answer." Someone
asks the right question — what state was the graph in when it gave up? — and
there is no answer, because the terminal that caught the failure is already
gone, the Kubernetes pod has restarted, and the LangSmith URL was never
recorded.
This skill produces one artifact: a single bundle-<incident_id>.tar.gz (typically
1-10 MB) containing everything a second engineer needs to reproduce the failure
without a live terminal — environment and version manifest, filtered
astream_events(version="v2") JSONL, a propagating callback stack, the
LangSmith trace URL, and a post-write sanitization pass.
Four pitfalls make naive bundles useless:
P01 — ChatAnthropic.stream() reports token_usage only on stream close; token math read from on_llm_end lags by stream duration, so cost context in the bundle is wrong.
P28 — BaseCallbackHandler.with_config(callbacks=[...]) does NOT propagate into subgraphs or inner create_react_agent loops. A debug callback bound that way silently captures zero events from the place the incident actually happened.
P47 — astream_events(version="v2") emits 2,000+ events per invocation. A raw dump is 50 MB and unreadable; an SSE viewer crashes on it.
P67 — astream_log() is soft-deprecated in 1.0. Diagnostic tooling built on it breaks on the next minor version.
The skill's answer: assemble the manifest, capture v2 events with a whitelist
(drop lifecycle noise, keep on_chat_model_stream / on_tool_* / any *_error
event), attach DebugCallbackHandler via config["callbacks"] at invoke time,
pull the LangSmith URL from the active RunTree, run the sanitization pass,
tar it up. Pinned: langchain-core 1.0.x, langgraph 1.0.x, langsmith 0.1.x.
Pain-catalog anchors: P01, P28, P47, P67.
Active LangSmith project (LANGSMITH_TRACING=true, LANGSMITH_API_KEY=...,
LANGSMITH_PROJECT=...) — canonical 1.0 env-var names, not the legacy
LANGCHAIN_TRACING_V2 (see P26).
Write access to a staging directory outside the repo tree.
Instructions
Step 1 — Assemble the environment manifest
Record the runtime snapshot that lets a colleague reproduce on a different
host. See env-manifest-template.md for
the exact YAML shape.
import platform, sys, os, subprocess, datetime
RELEVANT = [
"langchain-core", "langchain", "langgraph",
"langchain-anthropic", "langchain-openai",
"langsmith", "anthropic", "openai", "pydantic",
]
defpip_show(name: str) -> str | None:
try:
out = subprocess.check_output(
[sys.executable, "-m", "pip", "show", name],
stderr=subprocess.DEVNULL, text=True,
)
for line in out.splitlines():
if line.startswith("Version:"):
return line.split(":", 1)[1].strip()
except subprocess.CalledProcessError:
returnNonedefbuild_manifest(incident_id: str, invoke_meta: dict) -> dict:
return {
"bundle_spec_version": "1.0",
"generated_at": datetime.datetime.utcnow().isoformat() + "Z",
"incident_id": incident_id,
"runtime": {
"python": sys.version.split()[0],
"platform": platform.platform(),
"cpu_count": os.cpu_count(),
},
"packages": [
{"name": n, "version": pip_show(n)}
for n in RELEVANT if pip_show(n) isnotNone
],
# NAMES only — never values. Sanitized by design (P27 posture)."env_var_names_present": sorted(
k for k in os.environ
if k.startswith(("LANGSMITH_", "LANGCHAIN_", "ANTHROPIC_", "OPENAI_", "GOOGLE_"))
),
"invocation": invoke_meta,
}
Record env-var names, not values. Values go through the sanitization pass in
Step 5, but the safest design is never to capture them.
Step 2 — Capture astream_events(version="v2") with a filter
Raw v2 events flood 2,000+ per invocation (P47). A server-side filter drops
lifecycle noise (on_chain_start/on_chain_end) and keeps model, tool, and
error events — yielding 50-200 events per invocation and a ~500 KB JSONL.
import json, itertools
from pathlib import Path
KEEP = {
"on_chat_model_start", "on_chat_model_end",
"on_tool_start", "on_tool_end", "on_tool_error",
"on_retriever_start", "on_retriever_end",
"on_custom_event",
}
# Additionally: any event whose name ends in "_error"# Additionally: 1-in-10 sampled on_chat_model_stream (for response reconstruction)asyncdefcapture_events(graph, inputs, config, out_path: Path) -> int:
sample = itertools.count()
written = 0with out_path.open("w") as f:
asyncfor evt in graph.astream_events(inputs, config=config, version="v2"):
name = evt["event"]
if name == "on_chat_model_stream"andnext(sample) % 10 != 0:
continueif name notin KEEP andnot name.endswith("_error"):
continue
f.write(json.dumps({
"event": name,
"name": evt.get("name"),
"run_id": str(evt.get("run_id")),
"tags": evt.get("tags"),
"metadata": evt.get("metadata"),
"data": _json_safe(evt.get("data", {})),
}, default=str) + "\n")
written += 1return written
Never use astream_log() (P67). The full event taxonomy and _json_safe
helper live in astream-events-capture.md.
Step 3 — Attach callbacks that propagate into subgraphs (P28)
Callbacks bound via Runnable.with_config(callbacks=[...]) fire on the outer
chain only. They go silent the moment the graph crosses into a subgraph or an
inner create_react_agent loop — exactly where incidents happen. Pass them via
config["callbacks"] at invoke time instead.
The full handler (LLM + retriever + tool lifecycle) and a propagation smoke
test live in callback-propagation.md.
Step 4 — Record the LangSmith trace URL
A trace URL is cheaper than any local artifact — one click and the colleague
sees the full run with latency, token counts, and input/output per node. Pull
it from the active RunTree if you have a live handle; otherwise construct it
from the invoke's run_id:
from langsmith.run_helpers import get_current_run_tree
defcapture_langsmith_url() -> str | None:
rt = get_current_run_tree()
if rt isNone:
returnNone# tracing not enabled or run already closedreturn rt.get_url() # https://smith.langchain.com/o/.../r/<run_id># Write to langsmith.url in the bundle:
url = capture_langsmith_url()
(staging / "langsmith.url").write_text(url or"(no trace URL available)")
The URL requires the colleague to have access to the LangSmith project. For
public sharing, use RunTree.share() to generate a public snapshot URL. Never
paste a non-shared URL into a public Discord thread — the page redirects to a
login and leaks the project name.
Step 5 — Sanitize before packaging
Every file in the staging dir passes through the redaction pass before the
tar.gz is written. This is the last-mile guard; upstream redaction middleware
should already have caught credential material, but the bundle cannot assume
that.
import re
PATTERNS = [
("openai_key", r"sk-proj-[A-Za-z0-9_-]{16,}|sk-[A-Za-z0-9_-]{32,}"),
("anthropic_key", r"sk-ant-[A-Za-z0-9_-]{16,}"),
("google_key", r"AIza[A-Za-z0-9_-]{35}"),
("langsmith_key", r"lsv2_(?:pt|sk)_[A-Za-z0-9]{32,}"),
("bearer", r"(?i)bearer\s+[A-Za-z0-9._~+/=-]{20,}"),
("db_uri", r"[a-z]+://[^:/\s]+:[^@\s]+@[^/\s]+"),
("private_key", r"-----BEGIN [A-Z ]*PRIVATE KEY-----[\s\S]*?-----END [A-Z ]*PRIVATE KEY-----"),
]
defsanitize_file(path, patterns=PATTERNS) -> dict[str, int]:
text, counts = path.read_text(), {}
for name, pat in patterns:
new, n = re.subn(pat, f"[REDACTED:{name}]", text)
if n: counts[name] = n; text = new
path.write_text(text)
return counts
The full pattern catalog (credentials, session tokens, PII, internal URLs) and
the pre-upload tar -xzf ... && grep scan live in
sanitization-checklist.md. For the
production-grade upstream redaction middleware, use the forthcoming
langchain-security-basics skill.
Step 6 — Bundle with an index
Write a top-level MANIFEST.yaml that describes every file and records the
sanitization summary. Then tar.gz the staging dir.
Typical size: 1-10 MB compressed. Typical event count after filter: 50-200
per invocation (down from 2,000+ raw). Bundle is self-contained — no external
dependencies beyond tar -xzf and a text editor.
Error Handling
Error
Cause
Fix
events.jsonl has no subgraph events
Callbacks bound via Runnable.with_config(callbacks=[...]) instead of config["callbacks"] (P28)
Confirm langsmith.url is a shared URL (public), not a project-internal one
Strip the incident_id if it maps to internal ticket numbers you cannot disclose
Include in the post: bundle attachment, a 3-sentence symptom description, the
exact reproducer prompt, the first line of MANIFEST.yaml (spec version
and versions of langchain-core + langgraph)
