sawz/last30days-skill
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: --competitors flag for auto-discovered comparison fan-out (#308)

Browse Source 
Pass `--competitors` on a single-entity topic and the engine auto-discovers
2-6 peer entities via web search, runs the full pipeline on each in
parallel, and returns one N-way comparison reusing the existing 9-axis
Head-to-Head scaffold. `last30days OpenAI --competitors` resolves to
Anthropic + xAI + Google Gemini; `last30days Kanye West --competitors`
resolves to Drake + Kendrick Lamar + one more peer.

- New CLI flags: --competitors, --competitors=N, --competitors-list
- New scripts/lib/competitors.py — mirrors resolve.auto_resolve pattern
  (web search + deterministic text extraction, no internal LLM)
- New scripts/lib/fanout.py — ThreadPoolExecutor orchestrator; per-entity
  failures degrade gracefully as long as >=2 entities survive
- Multi-report render in scripts/lib/render.py reuses the comparison
  scaffold for the synthesis table
- LAW 7-style stderr when no backend and no list, pointing the hosting
  reasoning model at --competitors-list
- 38 new tests across CLI parsing, discovery, fanout, and rendering

Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Matt Van Horn
2026-04-22 21:28:36 -07:00
committed by GitHub
parent ff21243517
commit 5f054380c5
13 changed files with 1611 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CHANGELOG.md
+6
View File
@@ -5,6 +5,12 @@ All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## [3.0.11] - 2026-04-22
### Added
- **`--competitors` flag for auto-discovered comparison fan-out.** Pass `--competitors` on a single-entity topic and the engine discovers 2-6 peer entities via web search, then runs the full pipeline on each in parallel and emits one N-way comparison. `last30days Kanye West --competitors` resolves Drake, Kendrick Lamar, and one more peer. `last30days OpenAI --competitors` resolves Anthropic, xAI, Google Gemini. `--competitors=N` controls count, `--competitors-list="A,B,C"` skips discovery and uses the explicit list. Discovery mirrors the `auto_resolve` pattern (Brave / Exa / Serper / Parallel) with deterministic text extraction - no internal LLM call. Sub-runs inherit the main `--quick`/`--deep`/`--days`, run in a `ThreadPoolExecutor`, and degrade gracefully when at least 2 entities survive. Output reuses the existing 9-axis `## Head-to-Head` scaffold.
## [3.0.10] - 2026-04-21
### Added