111 lines
6.2 KiB
Markdown
111 lines
6.2 KiB
Markdown
<h1 align="center">Pocket Flow - LLM Framework in 100 Lines</h1>
|
||
|
||
|
||
|
||
|
||
[](https://the-pocket.github.io/PocketFlow/)
|
||
|
||
<div align="center">
|
||
<img src="./assets/minillmflow.jpg" width="400"/>
|
||
</div>
|
||
|
||
<br>
|
||
|
||
A [100-line](pocketflow/__init__.py) minimalist LLM framework for ([Multi-](https://the-pocket.github.io/PocketFlow/multi_agent.html))[Agents](https://the-pocket.github.io/PocketFlow/agent.html), [Task Decomposition](https://the-pocket.github.io/PocketFlow/decomp.html), [RAG](https://the-pocket.github.io/PocketFlow/rag.html), etc.
|
||
|
||
- Install via ```pip install pocketflow```, or just copy the [source codes](pocketflow/__init__.py) (only 100 lines)
|
||
|
||
- If the 100 lines feel terse and you’d prefer a friendlier intro, [check this out](https://chatgpt.com/share/678564bd-1ba4-8000-98e4-a6ffe363c1b8)
|
||
|
||
- **💡 Pro tip!!** Build LLM apps with LLMs assistants (ChatGPT, Claude, Cursor.ai, etc.)
|
||
|
||
<details>
|
||
<summary><b>(🫵 Click to expand) Use Claude to build LLM apps</b></summary>
|
||
|
||
- Create a [project](https://www.anthropic.com/news/projects) and upload the [docs](docs) to project knowledge
|
||
|
||
- Set project custom instructions. For example:
|
||
```
|
||
1. check "tool.md" and "llm.md" for the required functions.
|
||
2. design the high-level (batch) flow and nodes in artifact using mermaid
|
||
3. design the shared memory structure: define its fields, data structures, and how they will be updated.
|
||
Think out aloud for above first and ask users if your design makes sense.
|
||
4. Finally, implement. Start with simple, minimalistic codes without, for example, typing. Write the codes in artifact.
|
||
```
|
||
- Ask it to build LLM apps (Sonnet 3.5 strongly recommended)!
|
||
```
|
||
Help me build a chatbot based on a directory of PDFs.
|
||
```
|
||
|
||
<div align="center">
|
||
<img src="./assets/claude_project.gif"/>
|
||
</div>
|
||
</details>
|
||
|
||
<details>
|
||
<summary><b>(🫵 Click to expand) Use ChatGPT to build LLM apps</b></summary>
|
||
|
||
- Try the [GPT assistant](https://chatgpt.com/g/g-677464af36588191b9eba4901946557b-mini-llm-flow-assistant). However, it uses older models, which are good for explaining but not that good at coding.
|
||
|
||
<div align="center">
|
||
<img src="./assets/gpt_store.gif"/>
|
||
</div>
|
||
|
||
- For stronger coding capabilities, consider sending the [docs](docs) to more advanced models like O1.
|
||
|
||
- Paste the docs link (https://github.com/the-pocket/PocketFlow/tree/main/docs) to [Gitingest](https://gitingest.com/).
|
||
|
||
- Then, paste the generated contents into your O1 prompt, and ask it to build LLM apps.
|
||
|
||
|
||
</details>
|
||
|
||
|
||
Documentation: https://the-pocket.github.io/PocketFlow/
|
||
|
||
## Why Pocket Flow?
|
||
|
||
Pocket Flow is designed to be **the framework used by LLMs**. In the future, LLM projects will be *self-programmed* by LLMs themselves: Users specify requirements, and LLMs will design, build, and maintain. Current LLMs are:
|
||
|
||
1. **👍 Good at Low-level Details:** LLMs can handle details like *wrappers, tools, and prompts*, which don't belong in a framework. Current frameworks are over-engineered, making them hard for humans (and LLMs) to maintain.
|
||
|
||
2. **👎 Bad at High-level Paradigms:** While paradigms like *MapReduce, Task Decomposition, and Agents* are powerful, LLMs still struggle to design them elegantly. These high-level concepts should be emphasized in frameworks.
|
||
|
||
The ideal framework for LLMs should (1) **strip away low-level implementation details**, and (2) **keep high-level programming paradigms**. Hence, we provide this minimal (100-line) framework that allows LLMs to focus on what matters.
|
||
|
||
Pocket Flow is also a *learning resource*, as current frameworks abstract too much away.
|
||
|
||
| Framework | Computation Models | Communication Models | App-Specific Models | Vendor-Specific Models | Lines Of Codes | Package + Dependency Size |
|
||
|:--------------:|:------------------:|:--------------------:|:-------------------------------------------------------:|:--------------------------------------------------------:|:-----------------:|:---------------------------:|
|
||
| LangChain | Agent, Chain | Message | Many | Many | *405K* | *+166MB* |
|
||
| CrewAI | Agent, Chain | Message, Shared | Many | Many | *18K* | *+173MB* |
|
||
| SmolAgent | Agent | Message | Some | Some | *8K* | *+198MB* |
|
||
| LangGraph | Agent, Graph | Message, Shared | Some | Some | *37K* | *+51MB* |
|
||
| AutoGen | Agent | Message | Some | Many | *7K* | *+26MB* |
|
||
| **PocketFlow** | **Graph** | **Shared** | **None** | **None** | **100** | **+56KB** |
|
||
|
||
|
||
## How Does it Work?
|
||
|
||
The [100 lines](pocketflow/__init__.py) capture what we see as the core abstraction of most LLM frameworks: **Nested Directed Graph** that breaks down tasks into multiple (LLM) steps, with branching and recursion for agent-like decision-making.
|
||
|
||
<div align="center">
|
||
<img src="./assets/abstraction.png" width="500"/>
|
||
</div>
|
||
|
||
From there, it’s easy to layer on more complex features.
|
||
|
||
<div align="center">
|
||
<img src="./assets/paradigm.png" width="500"/>
|
||
</div>
|
||
|
||
- To learn more details, please check out documentation: https://the-pocket.github.io/PocketFlow/
|
||
|
||
- Beginner Tutorial: [Text summarization for Paul Graham Essay + QA agent](https://colab.research.google.com/github/the-pocket/PocketFlow/blob/main/cookbook/demo.ipynb)
|
||
|
||
- Have questions for this tutorial? Ask LLM assistants through [this prompt](https://chatgpt.com/share/676f16d2-7064-8000-b9d7-f6874346a6b5)
|
||
|
||
- More coming soon ... Let us know you’d love to see!
|
||
|
||
|