`go test -tags bench ./internal/tldr -run TestBench` runs the real /tldr
pipeline against one or more Ollama models and prompt variants on a chat dump
from disk (text copied from Discord, "Author: message" lines, or a
DiscordChatExporter JSON export) and writes a self-contained HTML dashboard
plus a JSON of the results. Scoring uses an optional facts file (a fact
passes only when one bullet holds all its details) and automatic checks:
numbers not in the chat, copying the chat, an English intro, cut-off output.
TestFactsSkeleton proposes candidate facts from the chat to prune by hand.
Chat dumps and results stay local: bench-out/ and the private folders are
git-ignored. See internal/tldr/tldrbench/README.md.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
/tldr [since] summarizes the messages in the channel it is run in that the
caller has not been summarized yet, using a small local model on Ollama, and
replies ephemerally. The answer streams into the reply as it is written.
- Per-user, per-channel cursor (tldr_cursors) so each run picks up where the
last left off; `since:` (e.g. 6h, 2d, max 7d) peeks without moving it.
First use covers the last 24h, and a stale cursor never reaches back more
than 7 days.
- Only summarizes channels both the caller and the bot can read.
- The transcript drops bots, system messages, reactions and filler, turns
links, GIFs, images and files into short tags ([link: site], [image],
[file: name], preview titles), marks replies, and writes a date marker
where the chat paused for 3h or more so the model can keep topics apart.
- The prompt asks for one bullet per conversation, oldest first, with a
worked example; it was tuned against a synthetic chat and a real export
with qwen3:4b-instruct-2507 (the default model).
- One model call at a time across the bot, one /tldr per user.
- docker-compose: Ollama service (CPU only, 8g memory cap: 6g killed the
runner mid-request), model pull on startup.
Needs the tldr_cursors migration (velox-dashboard-api 000016).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
FernandoJVideiraandClaude Opus 5.5