Agently
Agently 4 is a Python framework for building GenAI applications, installed with `pip install -U agently` and used with `from agently import Agently`. Its stated aim is to give GenAI application engineers the best developer experience by pulling LLM uncertainty back inside an engineering boundary. The README abstracts away varying model parameters, output formatting and the communication between engineering modules and GenAI logic, while leaving business logic to you. Its main idea is contract-first structured output: define a schema with `output()`, enforce critical fields with `ensure_keys` (including wildcard paths for list items) and let the framework parse and repair the result, without relying on provider-specific JSON-schema switches.
Models are configured through `OpenAICompatible` settings (base URL, model, auth), so local, hosted and proxy providers share one interface, with options such as `full_url` and custom auth headers. The README also covers tool planning with traceability, retries for missing fields and a path from demo to production, with examples under `examples/step_by_step/` including an auto-loop FastAPI service (SSE, WebSocket, POST, Docker-ready). Version 3 code lives on the `v3-final` branch. It is Apache-2.0 licensed with a docs site and English and Chinese introductions.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-13
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 6
- Latest release
- v4.1.4.10 (2026-10-13)
- Open issues
- 16
- Recent issues with at least one comment
- 10/12
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #342 Consolidate sandbox provider contributions #333–#337 / 收敛沙箱 Provider 贡献10 comments · 2026-07-31
- #324 Should local sandbox capabilities use system adaptation? 本地沙箱能力是否应该使用系统适配?7 comments · 2026-07-15
- #255 建议agent 支持skill5 comments · 2026-03-29
- #287 IndexError: list index out of range in response_adapter.py:164 when streaming API sends final chunk without choices5 comments · 2026-06-06
- #359 Provider suggestion: AnyLLM gateway — one key for the full Chinese model pool (DeepSeek, Qwen, GLM, Kimi, MiniMax, Doubao, ERNIE)2 comments · 2026-08-18
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Compatibility
Quick start
Use cases
Alternatives
Comments
No comments yet. Be the first!