Turn code into context. The AI-powered standup bot that translates raw git activity into stakeholder-ready updates.
Most standup bots are Vitamins: they just nag you to type what you did yesterday. Elapse is a Painkiller: it watches your code, analyzes the diffs, and writes the update for you.
It solves the Translation Gap:
- Dev types:
fix: update reducer logic - Manager reads: "What does this mean?"
- Elapse writes: "Fixed a race condition in the checkout flow, preventing double-charges."
- Diff-Aware Intelligence: It doesn't trust commit messages. It reads the code diffs to understand what actually changed.
- Daily & Weekly Reports: Get daily standups and executive-ready weekly rollups with RAG status indicators.
- Blocker Detection: Automatically surfaces PRs with
CHANGES_REQUESTEDreviews or stale review requests. - Zero Metrics: No commit counting. No velocity tracking. No spyware. Just visibility.
- Real-Time Ingestion: Processes webhooks instantly via Redis queues; no massive API bursts in the morning.
- One-Click Setup: No manual GitHub App configuration. Just deploy and click a button.
- Self-Hostable: Runs on your VPS with a single
docker compose up. - Model Agnostic: Supports Google Gemini (default), OpenAI, Anthropic via Vercel AI SDK.
git clone https://github.com/kdcokenny/elapse.git
cd elapse/deploy/docker
cp ../../.env.example .env
# Edit .env with your values
docker compose up -dVisit http://localhost:3000 to complete the GitHub App setup.
See deploy/docker/ for the full guide.
- Create a Docker Compose service in Dokploy
- Set compose file to
deploy/dokploy/compose.yml - Add environment variables
- Deploy and configure domain
See deploy/dokploy/ for the full guide.
| Variable | Description | Required |
|---|---|---|
GOOGLE_GENERATIVE_AI_API_KEY |
Your Gemini API key | Yes |
DISCORD_WEBHOOK_URL |
Default webhook for all reports | Yes |
PROJECT_CONTEXT |
Describes your product so AI understands business value | No |
TEAM_TIMEZONE |
Timezone for reports (IANA format) | No (default: America/New_York) |
LLM_MODEL_NAME |
AI model to use | No (default: gemini-flash-latest) |
LOG_LEVEL |
Logging verbosity | No (default: info) |
| Variable | Description | Default |
|---|---|---|
REPORT_CADENCE |
Report frequency: daily, weekly, or both |
weekly |
DAILY_SCHEDULE |
Cron for daily reports | 0 9 * * 1-5 (9 AM Mon-Fri) |
WEEKLY_SCHEDULE |
Cron for weekly reports | 0 16 * * 5 (4 PM Friday) |
Send daily and weekly reports to different Discord channels:
| Variable | Description |
|---|---|
DISCORD_WEBHOOK_URL |
Base webhook (used if overrides not set) |
DISCORD_WEBHOOK_URL_DAILY |
Override for daily reports |
DISCORD_WEBHOOK_URL_WEEKLY |
Override for weekly reports |
| Variable | Description | Default |
|---|---|---|
STALE_REVIEW_THRESHOLD_DAYS |
Days before a pending review is "stale" | 3 |
WEEKLY_RAG_BLOCKER_THRESHOLD |
Days before a blocker triggers 🔴 status | 7 |
GitHub App credentials (GITHUB_APP_ID, GITHUB_APP_PRIVATE_KEY, GITHUB_WEBHOOK_SECRET) are configured via the one-click setup wizard.
Elapse requires long-running processes and Redis. See the deploy/ directory for platform-specific guides:
| Platform | Guide |
|---|---|
| Docker | deploy/docker/ |
| Dokploy | deploy/dokploy/ |
More platforms may be added in the future.
Elapse uses an Ingest → Digest → Report pipeline:
GitHub Push → Probot → BullMQ → AI Translation → Redis → Reports → Discord
-
Ingest (Probot): Listens for GitHub push webhooks. Filters out bots, lockfiles, and merge commits. Pushes jobs to Redis queue.
-
Digest (Worker): Fetches the diff for each commit. Uses AI to translate changes into a single business-value sentence. Stores in Redis.
-
Report (Scheduler): Generates daily and/or weekly reports. Posts to Discord.
Generated at 9 AM on weekdays. Delivered as a color-coded embed in your Discord channel with full details in an auto-created thread. Includes:
- Narrative summary of the day's work
- Grouped by PR with translations
- Blocker callouts for PRs with
CHANGES_REQUESTEDor stale reviews - "Awaiting Review" section for pending reviews
Critical blockers (5+ days old) are surfaced directly in the main message for immediate visibility.
Executive-ready summaries generated on Fridays. Delivered as a compact, color-coded embed with the RAG status visible at a glance. Full breakdown is posted in an auto-created thread for drill-down.
Main message includes:
- RAG Status: 🟢 On Track / 🟡 At Risk / 🔴 Blocked (color-coded sidebar)
- Executive Summary: 1-2 sentence top-line
- Key Stats: PRs shipped, active blockers
- Escalations: Critical blockers (5+ days) surfaced directly
Thread contains:
- Shipped This Week: Thematically grouped PRs (3-5 groups)
- Blockers & Risks: Active blockers with age and mentions
- Help Needed: Escalations requiring action
- Carrying Into Next Week: In-progress work
RAG Status Logic:
| Status | Condition |
|---|---|
| 🟢 Green | No active blockers, < 3 stale reviews |
| 🟡 Yellow | Any active blocker, OR 3+ stale reviews |
| 🔴 Red | Any blocker ≥ 7 days, OR 3+ active blockers |
Weekend Handling: Commits pushed on Saturday/Sunday are attributed to Monday's daily report and the following week's weekly report.
# Install dependencies
bun install
# Run in development mode
bun run dev
# Run tests
bun test
# Lint and type check
bun run check- Runtime: Bun
- Framework: Probot
- Queue: BullMQ
- AI: Vercel AI SDK
- Logging: Pino
POST /api/github/webhooks- GitHub webhook receiverGET /health- Health check endpoint
Elapse automatically filters out noise:
- Bot commits: dependabot, renovate, github-actions
- Lockfile-only: package-lock.json, yarn.lock, etc.
- Merge commits:
Merge branch...,Merge pull request... - Vague messages:
fix,wip,update(AI relies on diff instead)
Elapse detects blockers automatically without requiring labels:
| Blocker Type | Detection Method |
|---|---|
| Changes Requested | PR has a review with CHANGES_REQUESTED state |
| Stale Review | Review requested but no response within threshold (default: 3 days) |
Blockers are surfaced in daily reports and aggregated in weekly reports with age tracking.
- Pre-existing PRs: Pull requests that were already open when you install Elapse will be tracked once they receive new commits. Full context (PR numbers, blockers) becomes complete within 1-2 weeks of normal development activity.
MIT