feat(027-checkpoint-resume-contrib): document checkpoint usage, storage location, and cleanup in README
This commit is contained in:
parent
a1a814ee77
commit
ee1a23cc70
44
README.md
44
README.md
|
|
@ -219,6 +219,50 @@ print(decision)
|
|||
|
||||
See `tradingagents/default_config.py` for all configuration options.
|
||||
|
||||
## Checkpoint / Resume
|
||||
|
||||
TradingAgents can save progress after each graph node so that a crashed or interrupted run resumes from the last successful step instead of starting over.
|
||||
|
||||
### Enabling Checkpoints
|
||||
|
||||
Via the CLI:
|
||||
```bash
|
||||
tradingagents analyze --checkpoint
|
||||
```
|
||||
|
||||
Via Python:
|
||||
```python
|
||||
config = DEFAULT_CONFIG.copy()
|
||||
config["checkpoint_enabled"] = True
|
||||
|
||||
ta = TradingAgentsGraph(debug=True, config=config)
|
||||
_, decision = ta.propagate("NVDA", "2026-01-15")
|
||||
```
|
||||
|
||||
When enabled, the graph logs `Resuming from step N` if a checkpoint is found for the same ticker + date, or `Starting fresh` otherwise. On successful completion the checkpoint is automatically cleared to avoid stale state.
|
||||
|
||||
### Storage Location
|
||||
|
||||
Checkpoints are stored as per-ticker SQLite databases under the data cache directory:
|
||||
|
||||
```
|
||||
~/.tradingagents/cache/checkpoints/<TICKER>.db
|
||||
```
|
||||
|
||||
Override the base path with the `TRADINGAGENTS_CACHE_DIR` environment variable.
|
||||
|
||||
### Cleanup
|
||||
|
||||
Delete all saved checkpoints before a run:
|
||||
```bash
|
||||
tradingagents analyze --clear-checkpoints
|
||||
```
|
||||
|
||||
Or remove them manually:
|
||||
```bash
|
||||
rm -rf ~/.tradingagents/cache/checkpoints/
|
||||
```
|
||||
|
||||
## Contributing
|
||||
|
||||
We welcome contributions from the community! Whether it's fixing a bug, improving documentation, or suggesting a new feature, your input helps make this project better. If you are interested in this line of research, please consider joining our open-source financial AI research community [Tauric Research](https://tauric.ai/).
|
||||
|
|
|
|||
Loading…
Reference in New Issue