Platform
Every surface, one brief, every score explained.
Topics, News, Repositories, Web intelligence, Alerts, and Bookmarks all read from the same brief, so what you see stays scoped to your thesis, and every rank comes with its reasons.
Surfaces
Detect, evaluate, act.
Detect
See what is moving
Topics
Trending clusters with real movement data: heat, burst ratio, and a trajectory over time, so you can tell a spike from a trend.
- Core repos versus today’s newcomers
- Trajectory view on 14, 30, or 90 days
- Pin the topics that matter to your thesis
Agent orchestration95.0
LLM observability87.7
Local inference73.1
News
Short, factual write-ups on what moved, ranked against your brief, each with a pitch before you commit to the full read.
- Pitch first, full article on demand
- Ranked to your brief, not a generic trending list
- Daily on Free, tighter cadence on paid plans
Pitch3 min read
LangGraph crosses a durable pull-request threshold
Contributor count and merged PRs both rose for six straight weeks, led by integrations rather than demos.
langchain-ai/langgraph
Evaluate
Understand what is real
Repositories
Every tracked repository ranked by relevance to your brief and grouped into Breakouts, Leaders, and Rising, with the signals behind each score.
- Growth and volume weighted for your profile
- Peers and related news on every repository
- Drill into recent activity in one click
langchain-ai/langgraphBreakout
huggingface/smolagentsRising
crewAIInc/crewAILeader
microsoft/autogenLeader
Web intelligence
Repository signals get outside context: discussion and adoption signals from beyond the repo itself, cross-checked against the activity.
- Web search integration for ecosystem context
- Cross-checks activity against outside discussion
- Adds context, never replaces the repository evidence
Repo activity
Web mentions
Added context
Act
Move when it matters
Alerts
A personal threshold derived from your interest axes. When something crosses it, it reaches you. Nothing below it does.
- Tune the threshold looser or tighter
- Every alert traces back to the axis that fired it
- No noise below your bar
— — Your thresholdCrossed
Bookmarks
Keep a running shortlist of repos, topics, and articles, separate from what the ranking surfaces for you automatically.
- Save any repo, topic, or article
- Independent of your automated ranking
- One place to come back to before a review
duckdb/duckdbRepo
Inference servingTopic
Vercel AI SDK v5 runtimeArticle
Methodology
How Dokeon reaches a score.
The sources, the computation, the freshness, and the editorial rules: everything you need to trust a rank, or to challenge it.
What we read
Repository activity (commits, pull requests, issues, contributors), releases and their cadence, and web and community discussion found through search. Your brief is the fifth input: it decides what matters.
How a score is built
Growth rate and absolute volume are measured separately, then combined with release cadence, outside discussion, and the historical trajectory. The weights follow the interest axes distilled from your brief, and the breakdown is shown next to every score.
How fresh it is
Rankings, topics, and momentum are recomputed every night. Each view shows the date of its snapshot, so you always know what you are looking at.
How we write
News pitches and briefs state observable facts and link back to their sources. Nothing is invented, nothing is exaggerated, and there is no manual curation queue between the signal and you.
Limits we state upfront
- GitHub activity is the backbone today. Web search adds context; it does not replace it.
- Open-source traction is not company quality. Dokeon tells you what is moving, not what to invest in.
- Examples on this site use real projects with illustrative values, never live claims.
Start with one brief. Decide with evidence.
Free on one market, no card required. Describe what you want to track and see what is really moving.