copy guide from notion

Author: LukeMainwaring

README.md

@@ -1 +1,114 @@
 # Agility Python API library
+
+# Installation
+
+You can install the Agility python client from [our internal repo](https://github.com/cleanlab/agility-python) with:
+
+```python
+pip install git+ssh://git@github.com/cleanlab/agility-python.git
+```
+
+# Overview
+
+This user guide goes over the basic concepts of Agility and includes an example of how to create an Agility assistant.
+
+Cleanlab Agility is an end-to-end API that allows you to create trustworthy AI applications. The API allows you to build reliable RAG systems by ingesting data from a wide variety of sources into Knowledge Bases and by creating AI Assistants that can respond to user queries while accessing the relevant data and providing reliable, interpretable answers. 
+
+## Concepts
+
+### Knowledge Bases
+
+- **Knowledge Bases** -
+- **Sources** -
+- **Documents** -
+
+### Assistants
+
+- **Assistant**  - An AI Assistant that will respond to user queries, while referencing user-supplied data sources and instructions.
+- **Threads -** A conversation session between a user and Assistant. Threads store Messages and retrieve context from Knowledge Bases to provide trustworthy results
+- **Messages -** A message created by a user or Assistant. Messages are represented as text and are attached to a parent Thread
+- **Runs -** Invokes the Assistant for a specific Thread. It uses the configuration for the Assistant as part of the context, including the data sources from its Knowledge Base and provided instructions. A Run will add Messages to the Thread
+
+
+# Quick Start
+
+## Step 0: Initialize python client
+
+```python
+from agility import Agility
+
+api_key = "your_agility_api_key"
+
+client = Agility(api_key=api_key)
+```
+
+## Step 1: Create a Knowledge Base
+
+```python
+knowledge_base = client.knowledge_bases.create(
+    name="Cleanlab docs",
+    description="Developer-facing docs for using Cleanlab",
+    ingestion_pipeline_params={
+        "curate": {},
+        "curate_document_store": {},
+        "transform": {},
+        "vector_store": {
+            "weaviate_collection_name": "weaviate_collection_name"
+        },
+    },
+)
+```
+
+## Step 2: Add a source to a Knowledge Base
+
+```python
+source = client.knowledge_bases.sources.create(
+    knowledge_base_id=knowledge_base.id,
+    name="Cleanlab docs URLs",
+    description="APIs and tutorials for Studio and TLM",
+    source_params={
+        "urls": ["https://help.cleanlab.ai", "https://help.cleanlab.ai/tlm"]
+    },
+    source_schedule={
+        "cron": "cron",
+        "utc_offset": 0,
+    },
+)
+```
+
+## ?? Step x: figure out if there are any other necessary actions related to sources and documents, such as source syncing
+
+## Step 3: Create an Assistant
+
+```python
+assistant = client.assistants.create(
+    knowledge_base_id=knowledge_base.id,
+    name="Chat with Cleanlab Docs",
+    description="Chat with Cleanlab Docs for Studio and TLM",
+)
+```
+
+## Step 4: Create a Thread
+
+```python
+thread = client.threads.create()
+```
+
+## Step 5: Add a Message to a Thread
+
+```python
+message = client.threads.messages.create(
+    thread_id=thread.id,
+    role="user",
+    content="I need to run TLM in batch-mode on a large number of datapoints. Can you help me?",
+)
+```
+
+## Step 6: Create a Run
+
+```python
+run = client.threads.runs.create(
+    thread_id=thread.id,
+    assistant_id=assistant.id,
+)
+```