Add version based support for agentic frameworks

[Aaditya-G]

Currently Agent Wiz displays what frameworks are supported, however it is not clear which versions of the framework is supported.

We need to add which latest version is supported for each framework.

[Aaryan-9]

Hey @Aaditya-G, please assign this to me. I’ll add latest supported versions to the CLI help for extract.

Approach:

• Add a small mapping (FRAMEWORKS_LATEST) in cli.py.
• Render it in agent-wiz extract --help

Please share latest tested versions for:

• agent_chat, autogen (core), crewai, google_adk, langgraph, llama_index, n8n, openai_agents, pydantic-ai, swarm

[Aaditya-G]

@Aaryan-9

The changes need to be made both in the CLI tool and in the README. it would be ideal to have a hook so that version support can be maintained at a single point, with both the CLI and README inferring from the hook.

Please share latest tested versions for:

agent_chat, autogen (core), crewai, google_adk, langgraph, llama_index, n8n, openai_agents, pydantic-ai, swarm

You can checkout the code in examples dir and cross check with official documentations to give us tested version.

[Jatkingmodern]

Worked with similar framework before kindly assign me this issue!

[Aaditya-G]

Approach ? @Jatkingmodern

[Jatkingmodern]

1. Latest tested versions (summary)

(These are the versions I could confirm from upstream package pages and repositories.)

agent_chat (AutoGen AgentChat): autogen-agentchat 0.7.5.
PyPI

autogen (core) (AutoGen core runtime): autogen-core 0.7.5.
PyPI

crewai: crewai 0.203.0 (PyPI release Oct 8, 2025).
PyPI

google_adk (Agent Development Kit / ADK): google-adk 1.16.0 (PyPI release Oct 9, 2025).
PyPI

langgraph: langgraph 0.6.10 (PyPI release Oct 9, 2025).
PyPI

llama_index: llama-index 0.14.4 (PyPI release Oct 3, 2025).
PyPI

n8n (node / repo release): n8n 1.115.2 (GitHub release/tag 2025-10-10; note: n8n is npm/Node, so pin tags).
GitHub

openai_agents (OpenAI Agents SDK): openai-agents 0.3.3 (GitHub release v0.3.3 Sept 30, 2025).
GitHub

pydantic-ai: (couldn't confirm a stable PyPI final version within the search budget) — please confirm the exact PyPI package name you want (there are similarly named libs). If you want, I can fetch the exact pypi/GitHub link and pin the value.
If you expect a specific package name (e.g. pydantic-ai, pydantic_ai or similar), tell me which one and I’ll pin it exactly.

swarm: the OpenAI educational framework openai/swarm is not a PyPI release; it’s recommended to install from the GitHub repo (or migrate to the OpenAI Agents SDK). If you want to pin to a specific commit I recommend using the recent commit referenced in the repo (e.g. b186492 / Mar 11, 2025) or install from git+https://github.com/openai/swarm.git.
GitHub
+1

2. FRAMEWORKS_LATEST mapping (suggestion)

Create a small module agent_wiz/framework_versions.py (or add into cli.py) and point both CLI and README to it.

agent_wiz/framework_versions.py

FRAMEWORKS_LATEST = {
# Python packages (pin exact pip install strings if you want)
"agent_chat": "autogen-agentchat==0.7.5", # AutoGen AgentChat (MS)
"autogen_core": "autogen-core==0.7.5", # AutoGen core (MS)
"crewai": "crewai==0.203.0",
"google_adk": "google-adk==1.16.0",
"langgraph": "langgraph==0.6.10",
"llama_index": "llama-index==0.14.4",
"openai_agents": "openai-agents==0.3.3",
"pydantic_ai": "pydantic-ai==", # <-- please confirm package name/version
# Non-PyPI / Node entries:
"n8n": "n8n@1.115.2", # (npm/GitHub tag)
# Swarm is GitHub-only (educational); recommend using the OpenAI Agents SDK for production
"swarm": "github:openai/swarm@b186492", # or "git+https://github.com/openai/swarm.git"
}

Note: I used the exact package strings for Python libs, and n8n + swarm as repo/tag strings; adjust to your repo conventions (e.g. just version numbers vs pip strings).

3. How to render it in agent-wiz extract --help (CLI snippet)

If cli.py uses argparse (or click), put a short help line that pulls from FRAMEWORKS_LATEST so you only maintain one mapping.

Example with argparse:

cli.py (excerpt)

from argparse import ArgumentParser
from agent_wiz.framework_versions import FRAMEWORKS_LATEST
from textwrap import indent

def pretty_versions_text():
items = [f"{k}: {v}" for k, v in FRAMEWORKS_LATEST.items()]
return "\n".join(items)

parser = ArgumentParser(
prog="agent-wiz",
description="Agent-Wiz - helpers & integrations"
)

include the mapping in the --help text for the relevant command

parser.add_argument(
"--show-supported-frameworks",
action="store_true",
help=f"Show latest supported framework versions (source: {name}::FRAMEWORKS_LATEST).\n\n{pretty_versions_text()}"
)

If you use click, set the help param similarly (or have a --supported subcommand that prints the mapping).

4. README integration (single-source hook)

Add a tiny code snippet to README.md that imports the values — this makes the README a living document, and keeps versions maintained in one place.

In README.md, under “Supported frameworks / CLI help”:

Supported frameworks (auto-sourced)

The CLI and docs read FRAMEWORKS_LATEST from agent_wiz/framework_versions.py. As of this commit:

• agent_chat: autogen-agentchat==0.7.5
• autogen_core: autogen-core==0.7.5
• crewai: crewai==0.203.0
• google_adk: google-adk==1.16.0
• langgraph: langgraph==0.6.10
• llama_index: llama-index==0.14.4
• n8n: n8n@1.115.2
• openai_agents: openai-agents==0.3.3
• swarm: install from GitHub (openai/swarm) or use OpenAI Agents SDK

Tip: to programmatically print this table for the README during CI, run a tiny Python script that imports FRAMEWORKS_LATEST and writes a markdown table into README.md (or a docs partial).

Implementing a CI step (optional): a small script that regenerates the README fragment from FRAMEWORKS_LATEST on every change, so README and CLI always match.

5. Notes, edge cases & suggestions

pydantic-ai: I couldn’t safely confirm the exact PyPI package name/version in this pass (there are multiple similarly named projects). Please confirm which package you mean (exact PyPI/GitHub link) and I’ll pin it precisely.

swarm: OpenAI’s swarm is an educational repo (install from GitHub). For production workflows, OpenAI recommends migrating to the OpenAI Agents SDK — I included both options in the mapping.
GitHub
+1

n8n is an npm/node project — pin by tag (github release) or in package.json if you vendor or run the n8n server.
GitHub

6. PR checklist (what I would commit)

Add agent_wiz/framework_versions.py with FRAMEWORKS_LATEST dictionary.

Update cli.py to import and show supported versions in --help (or add --show-supported-frameworks).

Update README.md to mention that the table is auto-sourced and show the current snapshot (or add a small generated fragment).

(Optional) Add a tiny scripts/update_readme_versions.py that regenerates README fragment from FRAMEWORKS_LATEST (run in CI on push).

Add a short unit test that imports FRAMEWORKS_LATEST and asserts each value is non-empty and follows expected pattern (e.g., == for PyPI pins or @ for npm/git entries).

Submit PR referencing the issue and include changelog notes.

Sources

AutoGen AgentChat (PyPI) — autogen-agentchat 0.7.5.
PyPI

AutoGen Core (PyPI) — autogen-core 0.7.5.
PyPI

CrewAI (PyPI) — crewai 0.203.0.
PyPI

Google ADK (PyPI) — google-adk 1.16.0.
PyPI

LangGraph (PyPI) — langgraph 0.6.10.
PyPI

LlamaIndex (PyPI) — llama-index 0.14.4.
PyPI

n8n (GitHub Releases) — n8n@1.115.2 (GitHub tag/release).
GitHub

OpenAI Agents SDK (GitHub / PyPI releases) — openai-agents v0.3.3.
GitHub

OpenAI Swarm (GitHub) — openai/swarm (install from GitHub or pin commit).

[Aaditya-G]

@Jatkingmodern we would really appreciate a human readable approach - this seems entirely generated from an AI system.

[Jatkingmodern]

well its well of the fact we are contributing to agentic AI, i respect the fact you are following code of condduct of contributing, kindly assign me this issue so that i can contribute with all my efforts

[Aaditya-G]

Thanks for your interest, but I won't be assigning you this issue right now.
My request for a human-readable approach was a necessary step to ensure alignment. Your response did not address that request and also indicated a fundamental misunderstanding of the project's domain - this project is about AI threat modeling and not agentic AI.
Successful contributions require a clear understanding of the project's goals. I'd recommend spending more time reading the code yourself and coming up with an approach that is not raw copy paste of an AI generated content.