📡 AI Dispatch

🇨🇳 中文版

Your daily AI intelligence briefing, delivered to Lark (Feishu).

Automatically aggregates the latest in AI, Robotics, and Agents every morning — analyzed by an LLM of your choice, published as a Lark cloud doc with a bot link notification. Runs entirely on GitHub Actions. No server. Cheap subscription with DeepSeek.

What You Get

Every digest contains five structured sections:

Section	Content
📌 Top Stories	10–15 curated items, each with significance analysis and cross-story connections
📈 Trend Analysis	Cross-article patterns with evidence and forward predictions
🔬 Papers Worth Reading	Selected arXiv papers with core contributions and reading focus
📖 Blog Pick	One deep-read recommendation (never repeats, auto-deduped)
💡 Today's Signal	The one judgment that matters most today, in one sentence

Quick Start

No terminal required — everything runs in your browser.

Prerequisites

GitHub account (free)
Lark (Feishu) app with im:message and docx:document permissions

Step 1 — Fork this repo

Click Fork in the top right → create it under your own account.

Step 2 — Run the Setup workflow

Go to Actions → ⚙️ Setup → Run workflow and fill in the form:

Field	What to enter
DeepSeek model	`deepseek-v4-flash` (default), `deepseek-v4-pro`, etc.
Output language	`English` or `中文`

The workflow updates config.yml and prints a checklist of the secrets you need to add next.

Step 3 — Add secrets

Go to Settings → Secrets and variables → Actions → New repository secret

Add these 5 secrets (the Setup workflow tells you exactly what to put in each):

Secret	Value
`DEEPSEEK_API_KEY`	API key from platform.deepseek.com/api_keys
`LARK_APP_ID`	App ID from open.feishu.cn/app
`LARK_SECRET`	App secret from the same Feishu app
`LARK_FOLDER_TOKEN`	Cloud folder token for doc storage (see docs/lark-doc.md)
`LARK_RECEIVER`	Recipient `union_id` (app needs `im:message` permission)

Optional — Langfuse LLM observability (see LLM Observability):

Secret	Value
`LANGFUSE_PUBLIC_KEY`	Project public key (`pk-lf-...`) from Langfuse
`LANGFUSE_SECRET_KEY`	Project secret key (`sk-lf-...`)
`LANGFUSE_BASE_URL`	Region host, e.g. `https://cloud.langfuse.com` or `https://jp.cloud.langfuse.com`

Step 4 — Verify

Go to Actions → ✅ Check Setup → Run workflow

── GitHub Secrets ──────────────────────────────────
  ✅  DEEPSEEK_API_KEY        (set)
  ✅  LARK_APP_ID             (set)
  ✅  LARK_SECRET             (set)
  ✅  LARK_RECEIVER           (set)
  ✅  LARK_FOLDER_TOKEN       (set)

── config.yml ──────────────────────────────────────
  ✅  config.yml found
  ✅  topics configured  (3 topics)
  ✅  news_feeds configured  (9 sources)
  ✅  blog_feeds configured  (8 blogs)

── DeepSeek API ─────────────────────────────────────
  ✅  API connection successful (deepseek-v4-flash)

── Lark ─────────────────────────────────────────────
  ✅  Lark configured
  ✅  Test Lark doc notification sent

══════════════════════════════════════════════════════
  🎉  All checks passed! Your daily digest starts tomorrow.
══════════════════════════════════════════════════════

Once all green, AI Dispatch runs automatically every day. The default schedule is UTC 6:00 (~07:00 BST) — change it in .github/workflows/daily_news.yml.

Prefer the command line?

Set up locally with the interactive wizard (requires Git, [uv](https://docs.astral.sh/uv/), and GitHub CLI).

Step 0 — Install Git, uv, and GitHub CLI

Install uv

# macOS / Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

Windows: Reopen your terminal after installation so uv is on your PATH.

Install Git

# macOS — comes pre-installed; if missing:
xcode-select --install

# Windows
winget install Git.Git

# Linux (Debian / Ubuntu)
sudo apt install git

Windows: After winget installs Git, close and reopen your terminal before continuing.

Install GitHub CLI

# macOS
brew install gh

# Windows — open a new terminal after this completes
winget install GitHub.cli

# Linux (Debian / Ubuntu)
sudo apt install gh

Windows: Same as above — reopen your terminal after installation so gh is on your PATH.

Log in to GitHub

gh auth login

Follow the prompts — select GitHub.com → HTTPS → Login with a web browser.

Step 1 — Fork, clone, and launch

# macOS / Linux
gh repo fork AkatQuas/ai-dispatch --clone
cd ai-dispatch        # use the folder name printed by gh above
uv sync
uv run python setup.py

# Windows
gh repo fork AkatQuas/ai-dispatch --clone
cd ai-dispatch        # use the folder name printed by gh above
uv sync
uv run python setup.py

gh prints the local path after cloning, e.g. Cloned fork's Git repository to ai-dispatch.

The wizard asks a few questions and handles everything else — secrets, config, and push.

Step 2 — Verify

Go to Actions → ✅ Check Setup → Run workflow and confirm all checks pass.

Cost

GitHub Actions is always free. The only cost is the DeepSeek API call for each daily digest (typically a few cents per run).

Model	Notes
`deepseek-v4-flash` / `deepseek-v4-pro`	Newer V4 models — see DeepSeek docs

Change the model in config.yml under digest.model, or pass it in the ⚙️ Setup workflow.

Local development

Dependencies are managed with uv (pyproject.toml + uv.lock). CI installs with uv sync --frozen.

cd ai-dispatch
uv sync                              # create .venv and install locked deps
uv sync --group dev                  # include ruff + pre-commit
uv run python setup.py               # interactive first-time setup
uv run python -m unittest tests.test_llm -v
uv run python check_setup.py             # full setup check (needs .env secrets)

Lint & format (Ruff):

uv run ruff check .                  # lint
uv run ruff check --fix .            # auto-fix
uv run ruff format .                 # format
uv run ruff format --check .         # CI: format check only

Git pre-commit (recommended):

uv sync --group dev
uv run pre-commit install            # once per machine
uv run pre-commit run --all-files    # manual full run

Each commit runs: trailing whitespace / EOF, YAML checks, Ruff lint (--fix) and format. See .pre-commit-config.yaml. CI mirrors this in Lint & Format workflow.

VS Code/Cursor: install the Ruff extension; .vscode/settings.json enables format-on-save.

CLI entry points:

Command	Description
`uv run ai-dispatch`	Fetch, summarize, and send daily digest
`uv run python check_setup.py`	Verify secrets, API, and Lark
`uv run ai-dispatch-issues`	Manage GitHub Issue state (`load` / `save` / …)

LLM Observability (Langfuse)

Langfuse is integrated as optional observability for every DeepSeek call. It is a practical way to see what the digest pipeline actually did — not just whether it succeeded.

When both LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY are set, each run records:

Traces for complete() and ping() — prompt in, model output out
Nested generations when a thinking model needs multiple rounds to finish
Layered trace hierarchy — summarize-digest → deepseek-complete → deepseek-round-N-{initial|reasoning|truncation} → deepseek-generation-N-*
Model, tokens, and cost per call — useful for comparing models and spotting regressions
Tags (ai-dispatch) for filtering in the Langfuse UI

If either key is missing, tracing is a no-op: no extra network calls, no behavior change.

Enable locally — add to .env (see .env.example):

LANGFUSE_PUBLIC_KEY=pk-lf-...
LANGFUSE_SECRET_KEY=sk-lf-...
LANGFUSE_BASE_URL=https://cloud.langfuse.com   # or https://jp.cloud.langfuse.com, etc.

Enable in CI — add the same three names as GitHub Actions secrets. The setup wizard (uv run python setup.py) can write them for you. Workflows already pass them through to LLM steps.

Create a free project at langfuse.com/cloud → Settings → API Keys.

File Structure

ai-dispatch/
├── config.yml              ← Your personalization (the only file to edit)
├── setup.py                ← Interactive setup wizard
├── check_setup.py          ← Setup verification (helper)
├── ai_dispatch/            ← Application library (single package layer)
│   ├── fetch_news.py       ← Main pipeline
│   ├── issue_store.py      ← Persist state/reports via GitHub Issues
│   ├── llm.py              ← DeepSeek API client
│   ├── langfuse_tracing.py ← Optional Langfuse tracing
│   ├── lark_doc.py         ← Lark cloud doc create + markdown write
│   ├── lark_notify.py      ← Doc report + bot link notification
│   ├── send_lark_message.py← Lark bot messaging
│   └── paths.py            ← Project root paths
├── tests/
│   └── test_llm.py
├── scripts/
│   └── smoke_test.py       ← pre-commit / CI smoke test
├── .pre-commit-config.yaml
├── pyproject.toml          ← Dependencies (managed with uv)
├── uv.lock
└── .github/workflows/
    ├── daily_news.yml      ← Daily cron job
    ├── lint.yml            ← Ruff lint / format / smoke test
    ├── setup.yml           ← First-time setup wizard (browser-based)
    └── check_setup.yml     ← One-click setup check

Runtime files (sent_history.json, report/*.md) are gitignored. CI loads/saves them through Issues (ai-dispatch-state, ai-dispatch-report). On each run: ① load pulls state + recent reports → ② fetch/analyze/send → ③ save + publish-today push back to Issues.

FAQ

Q: Check Setup passed but no daily digest? Check Actions → AI Dispatch for errors. GitHub Actions cron can occasionally delay 15–30 minutes.

Q: Lark message not received? Confirm LARK_RECEIVER is the recipient's union_id, LARK_FOLDER_TOKEN is set, and the app has im:message and docx:document permissions enabled.

Q: How do I change the output language? Edit output_language in config.yml. Default is English — change it to 中文 for Chinese output. The setup wizard also lets you choose during initial setup.

Q: How do I add my own RSS sources? Add a line under news_feeds or blog_feeds in config.yml: Source Name: https://rss-url.

Q: Blog picks keep repeating? Dedup state lives in the Issue labeled ai-dispatch-state. Clear the urls array in that Issue body (or locally in sent_history.json then run uv run ai-dispatch-issues save).

🇨🇳 中文版 → README.zh.md

Navigation Menu

Repository files navigation

📡 AI Dispatch

What You Get

Quick Start

Prerequisites

Step 1 — Fork this repo

Step 2 — Run the Setup workflow

Step 3 — Add secrets

Step 4 — Verify

Prefer the command line?

Step 0 — Install Git, uv, and GitHub CLI

Install uv

Install Git

Install GitHub CLI

Log in to GitHub

Step 1 — Fork, clone, and launch

Step 2 — Verify

Cost

Local development

LLM Observability (Langfuse)

File Structure

FAQ

About

Contributors

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 54 Commits
.github/workflows		.github/workflows
ai_dispatch		ai_dispatch
assets		assets
scripts		scripts
tests		tests
.env.example		.env.example
.gitignore		.gitignore
.pre-commit-config.yaml		.pre-commit-config.yaml
.python-version		.python-version
LICENSE		LICENSE
README.md		README.md
README.zh.md		README.zh.md
check_setup.py		check_setup.py
config.yml		config.yml
pyproject.toml		pyproject.toml
setup.py		setup.py
test.sh		test.sh
uv.lock		uv.lock

Folders and files

Latest commit

History

Repository files navigation

📡 AI Dispatch

What You Get

Quick Start

Prerequisites

Step 1 — Fork this repo

Step 2 — Run the Setup workflow

Step 3 — Add secrets

Step 4 — Verify

Prefer the command line?

Step 0 — Install Git, uv, and GitHub CLI

Install uv

Install Git

Install GitHub CLI

Log in to GitHub

Step 1 — Fork, clone, and launch

Step 2 — Verify

Cost

Local development

LLM Observability (Langfuse)

File Structure

FAQ

About

Resources

Stars

Watchers

Forks

Contributors

Languages