SKILL.md
Prepare a strategy for upload to OpenAlgo's /python strategy host.
Arguments
$0= strategy name (e.g.emacrossoverSBIN) OR full path to a.pyfile- Special:
--listlists deployable strategies instrategies/and their validation status
If $0 is a folder name, the file is strategies/<name>/strategy.py. If it's already a path, use as-is.
Instructions
- Read
algo-expert/rules/self-hosted-strategies.md. - Read the strategy file.
- Validate these criteria (each can be inferred via grep):
| Check | What to look for |
|---|---|
| SIGTERM handler | signal.signal(signal.SIGTERM, ...) present |
| SIGINT handler | signal.signal(signal.SIGINT, ...) present |
| stop_event used | threading.Event() and stop_event.set() |
| HOST_SERVER priority | os.getenv("HOSTSERVER") or os.getenv("OPENALGOHOST", pattern |
| Reads STRATEGY_EXCHANGE | os.getenv("OPENALGOSTRATEGYEXCHANGE", pattern |
| stdout logging | stream=sys.stdout in basicConfig; no FileHandler calls |
| Strategy entry guard | if name == "main": |
| Mode dispatcher | --mode argparse with default=os.getenv("MODE", "live") |
| No hardcoded local paths | /home/, C:\\Users, /Users/ not present |
| Env tag for STRATEGY_NAME | os.getenv("STRATEGY_NAME", pattern |
- Print a validation report:
`` Validation: strategies/emacrossoverSBIN/strategy.py [PASS] SIGTERM handler installed (line 327) [PASS] HOSTSERVER priority correct [PASS] Reads OPENALGOSTRATEGY_EXCHANGE [PASS] stdout-only logging [PASS] Mode dispatcher honors env MODE [WARN] No state.db persistence detected (optional) [PASS] No hardcoded local paths ALL REQUIRED CHECKS PASSED ``
- Generate
HOST_UPLOAD.mdin the same folder with the exact form values to enter athttp://localhost:5000/python:
# Upload Guide: <strategy_name>
## On http://localhost:5000/python
### Step 1: Click "Add Strategy"
| Form field | Value |
|---|---|
| Name | <strategy_name> |
| File | <full path to strategy.py> |
| Exchange | <SYMBOL's exchange, e.g. NSE> |
### Step 2: Parameters (key=value, one per row)
| Key | Value | Purpose |
|---|---|---|
| MODE | live | live execution (sandbox/real per UI analyzer toggle) |
| SYMBOL | <SYMBOL> | trading symbol |
| INTERVAL | <INTERVAL> | bar interval |
| LOG_LEVEL | INFO | logging verbosity |
### Step 3: Schedule
| Field | Value |
|---|---|
| Start time | 09:15 IST |
| Stop time | 15:30 IST (or leave empty for indefinite) |
| Days | Mon, Tue, Wed, Thu, Fri (weekdays for NSE) |
For MCX strategies use `EXCHANGE=MCX` and start time `17:00`.
For CRYPTO use `EXCHANGE=CRYPTO` and check all 7 days.
### Step 4: Click Upload, then Start
The host launches the strategy in a subprocess. Logs go to `openalgo/logs/strategies/<id>_<timestamp>.log`.
To stop the strategy gracefully, click Stop in the UI - the host sends SIGTERM and waits up to 15s for clean shutdown.
## Notes
- Live vs sandbox is decided by OpenAlgo UI's analyzer toggle, not by this strategy
- Re-upload anytime; the host preserves STRATEGY_ID
- The exchange you pick drives the calendar gating; intersect with start-stop- If
--list, liststrategies/subfolders, validate each, print a summary table.
When validation FAILs
Print the specific lines that need fixing:
[FAIL] HOST_SERVER priority: line 38 reads OPENALGO_HOST first.
Change:
API_HOST = os.getenv("OPENALGO_HOST", "http://127.0.0.1:5000")
To:
API_HOST = os.getenv("HOST_SERVER") or os.getenv("OPENALGO_HOST", "http://127.0.0.1:5000")Offer to make the fix automatically (use Edit tool).
Reset state
If the user asks to reset a strategy's persistent state before re-upload, suggest:
rm strategies/<name>/state.dbOr, if no state.db exists, no action needed.
Avoid
- Do not auto-upload via API - the
/pythonpage is a UI form, not an API endpoint - Do not modify the strategy's logic, only the host-compatibility plumbing
- Do not use icons/emojis