AgentAmi
AgentAmi is a flexible agentic framework built using LangGraph, designed to scale with large numbers of tools and intelligently select the most relevant ones for a given user query. It helps with decreasing token size significantly.
It supports:
- Dynamic tool selection via inbuilt runtime RAG (very efficient) with an option to easily replace it with your own tool_selector.
- Pruner to limit context length and improve performance (it's inbuilt, you don't have to do anything).
Quick start
Refer the main.py file for a complete sample usage.
pip install agentami
from agentami import AgentAmi
from langchain.chat_models import ChatOpenAI
from langgraph.checkpoint.memory import InMemorySaver
from agentami.agents.ami import AgentAmi
# Replace ... (ellipsis) with the commented instructions
tools = [...] # List of LangChain-compatible tools
agent = AgentAmi(
model=ChatOpenAI(model="gpt-4o"),
tools=tools, # List of LangChain-compatible tools
checkpointer=InMemorySaver(), # Optional. No persistence if omitted.
# Optional parameters:
tool_selector=..., # Custom function to select relevant tools. Defaults to internal tool_selector.
top_k=..., # Number of top tools to use. Defaults to 3.
context_size=..., # Number of past user prompts to retain. Defaults to 7.
disable_pruner=..., # If True, disables pruning & will increase token usage. Defaults to False
prompt_template=... # Custom prompt template. Defaults to a generic bot template.
)
agent_ami = agent.graph # Your regular langgraph's graph.
Things you should be aware about:
- Running for the first time will take time as it installs the dependencies (models used by internal tool_selector).
- Your first
agent_ami.invoke() or agent_agent_ami.astream()may take time if you have hundreds of tools, because it initialises a vector store and embeds the tool descriptions at runtime for each AgentAmi() object - Your eventual prompts' response time would be fine.
- Checkout ROADMAP.md file for future features.
How to integrate your own tool selector?
Just make a function that accepts (query: str, top_k: int) and parameters and returns List[str] #List of tool names.
from typing import List
# function template:
def my_own_tool_selector(query: str, top_k: int) -> List[str]:
# Your logic to select tools based on the query
return ["tool1", "tool2", "tool3"] # Return top_k selected tool names
Metadata
Release files for agentami 1.0.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| agentami-1.0.1.tar.gz | 9.0 kB | Without provenance | Details for agentami-1.0.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agentami-1.0.1-py3-none-any.whl | Python 3 | none | any | Details for agentami-1.0.1-py3-none-any.whl |
Total release size: 17.2 kB
Release files / agentami-1.0.1.tar.gz
| Download URL | agentami-1.0.1.tar.gz |
|---|---|
| Size | 9.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
b79b6e72975070dd9f0f5e0c82e8b20f46af0c4c941ccab22c2f1a99b029dc13
|
|
BLAKE2b-256 checksum How to use checksums |
daa27c18e8d2b701ca582e9983228165c565596fb5067699d7f64c44b9be3b27
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.12.9
|
Release files / agentami-1.0.1-py3-none-any.whl
| Download URL | agentami-1.0.1-py3-none-any.whl |
|---|---|
| Size | 8.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
0fd2e5607c3fed53a8d40b03988a5b5ddbc3a4182872922ed112c1caab2023d6
|
|
BLAKE2b-256 checksum How to use checksums |
51e827eeeccea5ed9cc47cdce0ebd6c12ad1083e54881b22df0ecaabc5a6c526
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.12.9
|
Security for agentami 1.0.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.