README describes LLM summarization that isn't wired in #4

Closed
opened 2026-07-10 15:39:24 +00:00 by Leonard · 2 comments

Category: Docs · Priority: MED

Summary
README says summarize.py "uses Ollama llama3.2:1b when present." Actual v1 is deterministic regex extraction (no LLM); the LLM path is a future --llm flag, not part of the daily cron. Ollama is installed but not running and no model is pulled.

Why it matters
Misleads operators about what's actually running; complicates debugging and erodes trust in the pipeline description.

What happens if not done
Future maintainers waste time expecting model output that never appears; misconfigured expectations around cost/latency.

Suggested approach
Update README to state v1 = deterministic extraction, LLM mode = future opt-in --llm flag, currently disabled. Reflect actual ENABLED_SOURCES.

**Category:** Docs · **Priority:** MED **Summary** README says `summarize.py` "uses Ollama llama3.2:1b when present." Actual v1 is deterministic regex extraction (no LLM); the LLM path is a future `--llm` flag, not part of the daily cron. Ollama is installed but not running and no model is pulled. **Why it matters** Misleads operators about what's actually running; complicates debugging and erodes trust in the pipeline description. **What happens if not done** Future maintainers waste time expecting model output that never appears; misconfigured expectations around cost/latency. **Suggested approach** Update README to state v1 = deterministic extraction, LLM mode = future opt-in `--llm` flag, currently disabled. Reflect actual `ENABLED_SOURCES`.
Author

Resolved. README updated to reflect actual state: v1 summarize.py = deterministic regex extraction (no LLM). Sub-agent summarization cron (athena-summarize-batch, every 4h) now provides LLM-quality summaries for high-signal entries, writing JSON back to entries.summary. Generator (generate_from_athena.py) wired to display one_liner under each headline on the live CFP feed.

Resolved. README updated to reflect actual state: v1 summarize.py = deterministic regex extraction (no LLM). Sub-agent summarization cron (`athena-summarize-batch`, every 4h) now provides LLM-quality summaries for high-signal entries, writing JSON back to `entries.summary`. Generator (`generate_from_athena.py`) wired to display `one_liner` under each headline on the live CFP feed.
Author

Resolved. README updated to reflect actual state: v1 summarize.py = deterministic regex extraction (no LLM). Sub-agent summarization cron (athena-summarize-batch, every 4h) now provides LLM-quality summaries for high-signal entries. Generator wired to display one_liner under each headline.

Resolved. README updated to reflect actual state: v1 summarize.py = deterministic regex extraction (no LLM). Sub-agent summarization cron (athena-summarize-batch, every 4h) now provides LLM-quality summaries for high-signal entries. Generator wired to display one_liner under each headline.
Sign in to join this conversation.
No Label
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Tony_tech/athena-oracle#4