Getting help
In shortWhere to ask questions, how to file a good bug report for 10xGraph, and what information to collect before you do.
- 2 min read
- 5 sections
- Updated
- v0.9.2
- Markdown
Start here
Most problems have a page already:
| Symptom | Page |
|---|---|
pip install fails, wrong Python version, agentflow not found |
Installation troubleshooting |
| Server will not start, 401/403, CORS, rate limits | API server troubleshooting |
| Streaming stalls, types missing, upload fails | Client troubleshooting |
| Playground shows nothing, or a page is empty | Playground troubleshooting |
You have an error code like GRAPH_ROUTING_001 |
Error codes |
| Something worked in 0.9 and broke after upgrading | Upgrade to 1.0 |
Search covers every page on this site. Press Ctrl + K.
Ask a question
- GitHub Discussions on 10xGraph/10xGraph for usage questions, design questions, and “is this the right approach” conversations.
- GitHub Issues on the repository for the package that misbehaves: the core library, the CLI and API server, or the TypeScript client.
For a security problem, do not use either. Follow the security policy.
File a good bug report
A report that includes the following usually gets a fix in one round trip instead of three.
# Versions of everything involved
agentflow version
python --version
pip show 10xgraph 10xscale-agentflow-cli | grep -E "Name|Version"
node --version && npm list @10xscale/agentflow-clientThen include:
- What you expected and what happened. The full traceback or the failing HTTP response body, not a paraphrase.
- A minimal graph that reproduces it. Strip the tools and prompts down to the smallest thing that still fails. Most bugs shrink to 20 lines.
- Your configuration. The relevant part of
10xgraph.json, which checkpointer and store you use, and whether you run withMODE=production. Redact secrets. - Whether it reproduces without persistence. Swapping
PgCheckpointerforInMemoryCheckpointerseparates a storage problem from a graph problem immediately.
If the run is long, enable structured logging and attach the records for the
failing run. They carry run_id, thread_id, and node, so a single run can be
extracted from a busy log:
from tenxgraph.utils.logging import setup_structured_logging
setup_structured_logging()Commercial support
10xGraph is built and maintained by 10xScale. For commercial support, architecture review, or production onboarding, contact [email protected].
Contributing a fix
Found the bug and know the fix? See contributing.