Spaces:
Runtime error
Runtime error
File size: 12,167 Bytes
0747ed8 cfc0db8 56724c7 8aaf6b2 ec49eb9 86c7913 386d73a 8aaf6b2 3234c2a 6ec321a 5ccaabd ec49eb9 386d73a 63081bb 6a8232f 5ccaabd 6a8232f 63081bb 6a8232f 784c769 5ccaabd cfc0db8 56724c7 507e0ff 6ec321a 507e0ff 56724c7 cfc0db8 56724c7 ec49eb9 86c7913 cfc0db8 51eee05 cfc0db8 51eee05 cfc0db8 30daf06 c768949 30daf06 c768949 56724c7 c768949 f47d3bd 3e355cf d07d5ac cfc0db8 d07d5ac cfc0db8 d07d5ac c768949 f47d3bd cfc0db8 f47d3bd ece814c f47d3bd 784c769 ece814c c768949 ece814c 784c769 56724c7 4283a03 ece814c 56724c7 ec49eb9 86c7913 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 |
---
title: π€ Megabots
emoji: π€
colorFrom: yellow
colorTo: green
sdk: gradio
sdk_version: 3.24.1
app_file: app.py
pinned: false
license: mit
python_version: 3.10.0
---
# π€ Megabots
[![Tests](https://github.com/momegas/qnabot/actions/workflows/python-package.yml/badge.svg)](https://github.com/momegas/qnabot/actions/workflows/python-package.yml)
[![Python Version](https://img.shields.io/badge/python-%203.10%20-blue.svg)](#supported-python-versions)
[![Code style: black](https://img.shields.io/badge/code%20style-black-000000.svg)](https://github.com/psf/black)
[![License](https://img.shields.io/badge/License-MIT-informational.svg)](https://github.com/momegas/megabots/blob/main/LICENCE)
![](https://dcbadge.vercel.app/api/server/zkqDWk5S7P?style=flat&n&compact=true)
π€ Megabots provides State-of-the-art, production ready LLM apps made mega-easy, so you don't have to build them from scratch π€― Create a bot, now π«΅
- π Join us on Discord: https://discord.gg/zkqDWk5S7P
- βοΈ Work is managed in this project: https://github.com/users/momegas/projects/5/views/2
- π€ Documentation bot: https://huggingface.co/spaces/momegas/megabots
**The Megabots library can be used to create bots that:**
- βοΈ are production ready, in minutes
- ποΈ can answer questions over documents
- πΎ can connect to vector databases
- ποΈ automatically expose the bot as a rebust API using FastAPI (early release)
- π automatically expose the bot as a UI using Gradio
**Coming soon:**
- π£οΈ accept voice as an input using [whisper](https://github.com/openai/whisper)
- π validate and correct the outputs of LLMs using [guardrails](https://github.com/ShreyaR/guardrails)
- π° semanticly cache LLM Queries and reduce Costs by 10x using [GPTCache](https://github.com/zilliztech/GPTCache)
- ποΈ mega-easy LLM training
- π mega-easy deployment
π€ Megabots is backed by some of the most famous tools for productionalising AI. It uses [LangChain](https://docs.langchain.com/docs/) for managing LLM chains, [FastAPI](https://fastapi.tiangolo.com/) to create a production ready API, [Gradio](https://gradio.app/) to create a UI. At the moment it uses [OpenAI](https://openai.com/) to generate answers, but we plan to support other LLMs in the future.
## Getting started
Note: This is a work in progress. The API might change.
```bash
pip install megabots
```
```python
from megabots import bot
import os
os.environ["OPENAI_API_KEY"] = "my key"
# Create a bot π with one line of code. Automatically loads your data from ./index or index.pkl.
# Keep in mind that you need to have one or another.
qnabot = bot("qna-over-docs")
# Ask a question
answer = bot.ask("How do I use this bot?")
# Save the index to save costs (GPT is used to create the index)
bot.save_index("index.pkl")
# Load the index from a previous run
qnabot = bot("qna-over-docs", index="./index.pkl")
# Or create the index from a directory of documents
qnabot = bot("qna-over-docs", index="./index")
# Change the model
qnabot = bot("qna-over-docs", model="text-davinci-003")
```
## Changing the bot's prompt
You can change the bots promnpt to customize it to your needs. In the `qna-over-docs` type of bot you will need to pass 2 variables for the `context` (knwoledge searched from the index) and the `question` (the human question).
```python
from megabots import bot
prompt = """
Use the following pieces of context to answer the question at the end.
If you don't know the answer, just say that you don't know, don't try to make up an answer.
Answer in the style of Tony Stark.
{context}
Question: {question}
Helpful humorous answer:"""
qnabot = bot("qna-over-docs", index="./index.pkl", prompt=prompt)
qnabot.ask("what was the first roster of the avengers?")
```
## Working with memory
You can easily add memory to your `bot` using the `memory` parameter. It accepts a string with the type of the memory to be used. This defaults to some sane dafaults.
Should you need more configuration, you can use the `memory` function and pass the type of memory and the configuration you need.
```python
from megabots import bot
qnabot = bot("qna-over-docs", index="./index.pkl", memory="conversation-buffer")
print(qnabot.ask("who is iron man?"))
print(qnabot.ask("was he in the first roster?"))
# Bot should understand who "he" refers to.
```
Or using the `memory`factory function
```python
from megabots import bot, memory
mem("conversation-buffer-window", k=5)
qnabot = bot("qna-over-docs", index="./index.pkl", memory=mem)
print(qnabot.ask("who is iron man?"))
print(qnabot.ask("was he in the first roster?"))
```
NOTE: For the `qna-over-docs` bot, when using memory and passing your custom prompt, it is important to remember to pass one more variable to your custom prompt to facilitate for chat history. The variable name is `history`.
```python
from megabots import bot
prompt = """
Use the following pieces of context to answer the question at the end.
If you don't know the answer, just say that you don't know, don't try to make up an answer.
{context}
{history}
Human: {question}
AI:"""
qnabot = bot("qna-over-docs", prompt=prompt, index="./index.pkl", memory="conversation-buffer")
print(qnabot.ask("who is iron man?"))
print(qnabot.ask("was he in the first roster?"))
```
## Using Megabots with Milvus (more DBs comming soon)
Megabots `bot` can also use Milvus as a backend for its search engine. You can find an example of how to do it below.
In order to run Milvus you need to follow [this guide](https://milvus.io/docs/example_code.md) to download a docker compose file and run it.
The command is:
```bash
wget https://raw.githubusercontent.com/milvus-io/pymilvus/v2.2.7/examples/hello_milvus.py
```
You can then [install Attu](https://milvus.io/docs/attu_install-docker.md) as a management tool for Milvus
```python
from megabots import bot
# Attach a vectorstore by passing the name of the database. Default port for milvus is 19530 and default host is localhost
# Point it to your files directory so that it can index the files and add them to the vectorstore
bot = bot("qna-over-docs", index="./examples/files/", vectorstore="milvus")
bot.ask("what was the first roster of the avengers?")
```
Or use the `vectorstore` factory function for more customisation
```python
from megabots import bot, vectorstore
milvus = vectorstore("milvus", host="localhost", port=19530)
bot = bot("qna-over-docs", index="./examples/files/", vectorstore=milvus)
```
## Exposing an API with FastAPI
You can also create a FastAPI app that will expose the bot as an API using the create_app function.
Assuming you file is called `main.py` run `uvicorn main:app --reload` to run the API locally.
You should then be able to visit `http://localhost:8000/docs` to see the API documentation.
```python
from megabots import bot, create_api
app = create_app(bot("qna-over-docs"))
```
## Exposing a Gradio chat-like interface
You can expose a gradio UI for the bot using `create_interface` function.
Assuming your file is called `ui.py` run `gradio qnabot/ui.py` to run the UI locally.
You should then be able to visit `http://127.0.0.1:7860` to see the API documentation.
```python
from megabots import bot, create_interface
demo = create_interface(bot("qna-over-docs"))
```
## Customising bot
The `bot` function should serve as the starting point for creating and customising your bot. Below is a list of the available arguments in `bot`.
| Argument | Description |
| ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| task | The type of bot to create. Available options: `qna-over-docs`. More comming soon |
| index | Specifies the index to use for the bot. It can either be a saved index file (e.g., `index.pkl`) or a directory of documents (e.g., `./index`). In the case of the directory the index will be automatically created. If no index is specified `bot` will look for `index.pkl` or `./index` |
| model | The name of the model to use for the bot. You can specify a different model by providing its name, like "text-davinci-003". Supported models: `gpt-3.5-turbo` (default),`text-davinci-003` More comming soon. |
| prompt | A string template for the prompt, which defines the format of the question and context passed to the model. The template should include placeholder variables like so: `context`, `{question}` and in the case of using memory `history`. |
| memory | The type of memory to be used by the bot. Can be a string with the type of the memory or you can use `memory` factory function. Supported memories: `conversation-buffer`, `conversation-buffer-window` |
| vectorstore | The vectorstore to be used for the index. Can be a string with the name of the databse or you can use `vectorstore` factory function. Supported DBs: `milvus`. |
| sources | When `sources` is `True` the bot will also include sources in the response. A known [issue](https://github.com/hwchase17/langchain/issues/2858) exists, where if you pass a custom prompt with sources the code breaks. |
## How QnA bot works
Large language models (LLMs) are powerful, but they can't answer questions about documents they haven't seen. If you want to use an LLM to answer questions about documents it was not trained on, you have to give it information about those documents. To solve this, we use "retrieval augmented generation."
In simple terms, when you have a question, you first search for relevant documents. Then, you give the documents and the question to the language model to generate an answer. To make this work, you need your documents in a searchable format (an index). This process involves two main steps: (1) preparing your documents for easy querying, and (2) using the retrieval augmented generation method.
`qna-over-docs` uses FAISS to create an index of documents and GPT to generate answers.
```mermaid
sequenceDiagram
actor User
participant API
participant LLM
participant Vectorstore
participant IngestionEngine
participant DataLake
autonumber
Note over API, DataLake: Ingestion phase
loop Every X time
IngestionEngine ->> DataLake: Load documents
DataLake -->> IngestionEngine: Return data
IngestionEngine -->> IngestionEngine: Split documents and Create embeddings
IngestionEngine ->> Vectorstore: Store documents and embeddings
end
Note over API, DataLake: Generation phase
User ->> API: Receive user question
API ->> Vectorstore: Lookup documents in the index relevant to the question
API ->> API: Construct a prompt from the question and any relevant documents
API ->> LLM: Pass the prompt to the model
LLM -->> API: Get response from model
API -->> User: Return response
```
## How to contribute?
We welcome any suggestions, problem reports, and contributions!
For any changes you would like to make to this project, we invite you to submit an [issue](https://github.com/momegas/megabots/issues).
For more information, see [`CONTRIBUTING`](https://github.com/momegas/megabots/blob/main/CONTRIBUTING.md) instructions.
|