Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
An open-source LLM based automatically daily news collecting workflow showcase powered by Agently AI application development framework.
| Date | Stars |
|---|---|
| 2026-08-02 | 635 |
| 2026-08-06 | 635 |
| 2026-08-12 | 635 |
| 2026-08-19 | 637 |
| 2026-08-20 | 637 |
| 2026-08-22 | 638 |
| 2026-08-24 | 638 |
| 2026-08-25 | 639 |
| 2026-08-26 | 639 |
| 2026-09-02 | 639 |
| 2026-09-06 | 640 |
| 2026-09-17 | 641 |
| 2026-09-26 | 642 |
| 2026-10-05 | 642 |
| 2026-10-14 | 643 |
| 2026-10-21 | 642 |
| 2026-10-22 | 643 |
| 2026-10-23 | 644 |
| 2026-10-26 | 644 |
| 2026-11-04 | 644 |
Today
— stars today
This week
— stars this week
This month
+2 stars this month
Momentum
0.0
growth rate 0.00%/day
# Agently Daily News Collector v4
Agently Daily News Collector has been rewritten on top of **Agently v4** and now uses:
- `TriggerFlow` for the end-to-end pipeline
- Agently v4 built-in `Search` and `Browse` tools
- structured output contracts instead of the old v3 workflow API
> Version constraint: this project requires **Agently v4.0.8.3 or newer**. The current implementation uses `TriggerFlow sub flow` to organize per-column pipelines, so earlier v4 releases are not compatible with the workflow structure used here.
The previous Agently v3 project has been archived under [`./v3`](./v3).
## Features
- Input a topic and generate a multi-column news briefing automatically
- Search, shortlist, browse, summarize, and assemble stories in one flow
- Save the final report as Markdown under `./outputs`
- Keep prompt templates in `./prompts` for easy editing
- Keep an independent `./tools` layer so search/browse can be replaced without touching the main workflow
- Keep flow construction in `./workflow` so orchestration can evolve independently from collector logic
## Quick Start
1. Install dependencies:
```bash
pip install -r requirements.txt
```
If you install Agently manually, make sure you use at least:
```bash
pip install "agently>=4.0.8.3"
```
2. Edit [`SETTINGS.yaml`](./SETTINGS.yaml):
- Keep the model block as environment placeholders
- Export the required environment variables:
```bash
export AGENTLY_NEWS_BASE_URL="https://api.openai.com/v1"
export AGENTLY_NEWS_MODEL="gpt-4.1-mini"
export AGENTLY_NEWS_API_KEY="your_api_key"
```
- Or put them in a local `.env` file:
```dotenv
AGENTLY_NEWS_BASE_URL=https://api.openai.com/v1
AGENTLY_NEWS_MODEL=gpt-4.1-mini
AGENTLY_NEWS_API_KEY=your_api_key
```
- Adjust language / search / concurrency settings if needed
- If your OpenAI-compatible endpoint does not require authentication, you can leave `AGENTLY_NEWS_API_KEY` unset and the project will skip `auth`.
3. Run:
```bash
python app.py
```
Or pass a topic directly:
```bash
python app.py "AI agents"
```
## Project Structure
```text
.
├── app.py
├── news_collector/
├── tools/
├── workflow/
├── prompts/
├── outputs/
├── logs/
└── v3/
```
## Important v3 -> v4 Changes
The business chain is still roughly:
`outline -> search -> pick -> browse + summarize -> write column -> render markdown`
What changed is the engineering shape around that chain.
### Project-level changes
- The old v3 project used a main workflow plus a nested column workflow under `./workflows`, with custom `search.py` / `browse.py` helpers and storage-style state passing.
- The v4 project separates responsibilities more clearly:
- `news_collector/`: app/integration layer
- `workflow/`: parent flow, column sub flow, and concrete chunk logic
- `tools/`: search/browse adapter layer
- `prompts/`: structured prompt contracts
- Model configuration is no longer hardcoded in Python. It now uses `${ENV.xxx}` placeholders from `SETTINGS.yaml`, so deployment and local switching are simpler.
- Tool wiring is no longer buried inside workflow code. Search, browse, and logger are injected as TriggerFlow runtime resources, which makes the workflow easier to replace or test.
- The workflow plan is now closer to the business boundary:
- parent flow: `prepare_request -> generate_outline -> for_each(column) -> render_report`
- column sub flow: `search -> pick -> summarize -> write_column`
- the `summarize` stage inside the column flow is further pushed down into a summary sub flow, where TriggerFlow handles fan-out and collection directly instead of leaving `asyncio.gather` in business code
- this keeps the parent focused on report orchestration and the child focused on one column lifecycle
- the immediate value of `sub flow` here is that the column pipeline becomes a reusable, independently evolvable workflow unit instead of staying buried inside one oversized parent chunk
### Agently v4 features used here
- **TriggerFlow orchestration**
- ReplaExcerpt of 8,466 characters
Read on GitHub
32
Erjan K · Netherlands
1
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:7fdeeedb776cb508, topic:llm-agent