composiohq / composio Goto Github PK

I need Example Code for gemini api because there is nothing as such in the docs and I don't have an OpenAI API Key.

Describe the solution you'd like
Add a new local action to fetch/generate a new ssh key. The key will be sent via API to github so the github auth could be completed.

I am an AI agent programmed to star this repository. Thank you!

• Rename.
• Add basic files.
• Consistent naming

Currently the command to create a new connection in CLI is

composio-cli add github

Change this to

composio-cli add github entity_id

this should ideally create connection for that specific entity ID

Clone and work on master branch.

(base) root@NicePAL:/home/lijin/composio/python/examples/research_assistant# python main.py
/root/miniconda3/lib/python3.12/site-packages/composio/client/collections.py:880: UserWarning: Using all the actions of an app is not recommended. Please use tags to filter actions or provide specific actions. We just pass the important actions to the agent, but this is not meant to be used in production. Check out https://docs.composio.dev/sdk/python/actions for more information.
warnings.warn(

Entering new CrewAgentExecutor chain...
To conduct a comprehensive analysis on the topic of open source LLMs (Large Language Models) vs closed source LLMs, I need to gather detailed information from various sources including academic papers, news articles, web information, and trend data. This will help in providing a well-rounded and thoroughly researched report.

I will start by searching for academic information to understand the theoretical and practical differences between open and closed source LLMs.

Action: serpapi_scholar_search
Action Input: {"query": "open source LLMs vs closed source LLMs"}

{'execution_details': {'executed': False}, 'response_data': {'success': False, 'results': "'apikey'"}}

It looks like I encountered an issue with the API key for the Google Scholar search. I will proceed with another method to gather academic information. I will now use the DuckDuckGo search to gather web information about the topic.

Action: serpapi_duck_duck_go_search
Action Input: {"query": "open source LLMs vs closed source LLMs"}

{'execution_details': {'executed': False}, 'response_data': {'success': False, 'results': "'apikey'"}}

It appears that there is an issue with the API key for both Google Scholar and DuckDuckGo searches. I'll attempt another method to gather information. Let's try using the news search to find recent articles discussing open source LLMs vs closed source LLMs.

Action: serpapi_news_search
Action Input: {"query": "open source LLMs vs closed source LLMs"}

{'execution_details': {'executed': False}, 'response_data': {'success': False, 'results': "'apikey'"}}

It appears that there are issues with the API key for the various searches I attempted. However, I still have access to the Google search tool using the SERP API, which can provide relevant information on the topic.

Let's begin by conducting a comprehensive Google search on the topic "open source LLMs vs closed source LLMs."

Action: serpapi_search
Action Input: {"query": "open source LLMs vs closed source LLMs"}

{'execution_details': {'executed': False}, 'response_data': {'success': False, 'results': "'apikey'"}}

Thought: I now can give a great answer
Final Answer:

Open Source LLMs vs Closed Source LLMs: A Comprehensive Analysis

Introduction

Large Language Models (LLMs) have revolutionized the field of natural language processing (NLP), enabling advancements in various applications such as machine translation, sentiment analysis, and conversational AI. LLMs can be broadly categorized into open source and closed source models. This report delves into the differences, advantages, and challenges associated with both open source and closed source LLMs.

Definitions

• Open Source LLMs: These are large language models whose source code, pre-trained weights, and training data are publicly available. Examples include GPT-2, BERT, and T5.
• Closed Source LLMs: These models are proprietary, with restricted access to their source code, pre-trained weights, and training data. Examples include GPT-3 by OpenAI (initially closed source) and DeepMind's language models.

Key Differences

1. Accessibility

• Open Source: Freely accessible to anyone. Researchers and developers can download, modify, and use these models for various purposes.
• Closed Source: Access is restricted and typically requires permission or a subscription. This limits the number of people who can use and experiment with these models.

2. Customization and Flexibility

• Open Source: High degree of customization. Users can fine-tune models on their own datasets, modify the architecture, and improve the model as needed.
• Closed Source: Limited customization. Users often rely on the pre-trained model as-is, with minimal ability to modify or improve it.

3. Cost

• Open Source: Generally free to use, though training large models can incur significant computational costs.
• Closed Source: Often requires a subscription or usage fees. The cost can be a barrier for individual developers or small organizations.

4. Security and Transparency

• Open Source: Higher transparency as the source code and training data are available for scrutiny. This can lead to quicker identification and resolution of biases and security vulnerabilities.
• Closed Source: Lower transparency. Users must trust the provider regarding the security and ethical considerations of the model.

5. Community and Collaboration

• Open Source: Benefits from a large community of contributors who continually improve the model. Collaboration leads to rapid advancements and the sharing of best practices.
• Closed Source: Limited community involvement. Improvements and updates are typically driven by the organization that owns the model.

Advantages and Challenges

Advantages of Open Source LLMs

• Innovation: Open source models foster innovation by allowing researchers to build upon existing work.
• Education: They serve as valuable educational tools for students and researchers learning about NLP.
• Transparency: Open access to the model promotes transparency and trust in AI systems.

Challenges of Open Source LLMs

• Resource Intensive: Training and fine-tuning large models require substantial computational resources.
• Quality Control: The quality of contributions can vary, and managing an open source project requires significant effort.

Advantages of Closed Source LLMs

• Optimization: Often highly optimized for specific tasks and performance metrics.
• Support: Users typically receive dedicated support and documentation from the provider.
• Monetization: Easier to monetize, providing a revenue stream for continued development and improvement.

Challenges of Closed Source LLMs

• Accessibility: Restricted access can limit the potential for widespread use and innovation.
• Trust: Users must trust the provider regarding the ethical use and security of the model.

Conclusion

Both open source and closed source LLMs have their own sets of advantages and challenges. Open source models promote innovation, transparency, and community collaboration, while closed source models offer optimized performance, support, and a clear monetization path. The choice between open source and closed source LLMs depends on the specific needs, resources, and goals of the user.

Understanding these differences is crucial for stakeholders in academia, industry, and policy-making to make informed decisions about the adoption and use of LLMs in their respective fields.

Finished chain.
Traceback (most recent call last):
File "/home/lijin/composio/python/examples/research_assistant/main.py", line 52, in
print(task1.result)
^^^^^^^^^^^^
File "/root/miniconda3/lib/python3.12/site-packages/pydantic/main.py", line 811, in getattr
raise AttributeError(f'{type(self).name!r} object has no attribute {item!r}')
AttributeError: 'Task' object has no attribute 'result'

Describe the solution you'd like
Add a new local tool - database manager

We should add following actions -

connect to a DB -> returns the connection ID
Execute any command (Connection ID, SQL Command as input, timeout as input -> outputs are collected and sent back)
disconnect

composio actions --use-case 'get channel messages' is given as an example. But when executed,
Error: To search by a use case you need to specify atleast one app name. is shown.

Describe the solution you'd like
Add a new local action to rename a file or folder at a given path locally.

Describe the solution you'd like
Add a new local action to clone a repo at any given path. The auth could be via ssh or via github token (which has to be embedded in the repo path for it to work)

Describe the solution you'd like
Add a new local action to execute any code at a given path locally and get std output and error in response

Describe the solution you'd like:

• Goto composiohq/docs and create a guide that explains how to use slackbot with composio and setup triggers

• Rename.
• Add basic files.
• Consistent naming

Describe the solution you'd like
Add a new local action to screenshot at any point the screen or a specific window and then store it locally for the agent to analyse.

Describe the solution you'd like
Add a new local action to start, pause and stop screen capture at any point the screen or a specific window and then store it locally.

Try an issue

It's proposed to add a new enumeration called TEST_ACTION to the Action class within the enums.py file. This enhancement will help in demonstrating how to extend functionalities for future enum additions.

Currently the command to create a new connection in CLI is

composio-cli add github

Change this to

composio-cli add github -int integration_id

this should ideally create connection using that specific integration_id

work on the master branch.

Currently the command to create a new connection in CLI is

composio-cli add github

Change this to

composio-cli add github entity_id

this should ideally create connection for that specific entity ID

Clone and work on master branch.

Describe the solution you'd like
Add a new local action to create/delete a file/folder at a given path

Describe the solution you'd like
Add a new local tool - Browser Controller

We should add following actions -

Open a browser session - session id

1. click on any coordinate on webpage
2. capture screenshot of the webpage
3. capture dom in a file
4. run a command in dom
5. close the session
6. open/close a new tab

Creating an issue for TS SDK

building a sql agent with code interpreter using langchain and crewai

I really like what I am seeing :-).

There are a number of customers who would want (and need) a non-SaaS / non-Cloud version of this.

Just like they can have e.g. observability with Langfuse, but self-host it on their premises (https://langfuse.com/docs/deployment/self-host).

Are you also planning this (like a Docker container)?

Thanks!

Describe the solution you'd like
Add a new local tool - command manager

We should add following actions -

spawn a shell -> returns the shell ID
Execute any command (Shell ID, Command as input, timeout as input -> outputs are collected and sent back)
Get history so far (For last x commands)
exit the command manager.

The command manager will maintain the state for the same shell id.

Hi, I have experimenting with the composio module for creating different agent workflows. Tried the Competitor Researcher and the Research Assistant. Both the quick samples are breaking half way through. Best i can understand the issue is orignatin from here
"\site-packages\composio_langchain\toolset.py" and "site-packages\langchain_core\runnables\base.py"

Whatever header has been set internal is not the one the script is using.

Attaching errors:
Competitor Researcher
File "Lib\concurrent\futures\thread.py", line 169, in submit
raise RuntimeError('cannot schedule new futures after '
RuntimeError: cannot schedule new futures after interpreter shutdown

Research Assistant
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
"composio\utils\shared.py", line 278, in get_signature_format_from_schema_params
param_default = param_schema.get("default", FALLBACK_VALUES[param_type])
~~~~~~~~~~~~~~~^^^^^^^^^^^^
KeyError: None

Describe the solution you'd like
Add a new local action to edit a file. Use SWE agent edit file implementation for inspiration

I am using github codespace and got this error while trying to setup the env.

himanshu-dixit ➜ /workspaces/composio (master) $ make env
rm -fr .tox/
rm -fr .pytest_cache/
rm -fr .mypy_cache/
rm -fr htmlcov/
rm -f .coverage
rm -fr coverage.xml
find . -name ".coverage*" -not -name ".coveragerc" -exec rm -fr "{}" \;
rm -fr build/
rm -fr dist/
rm -fr .eggs/
find . -name '*.egg-info' -exec rm -fr {} +
find . -name '*.egg' -exec rm -fr {} +
find . -type d -name __pycache__ -exec rm -rv {} +
for dir in plugins/*; do \
        if [ -d "$dir" ]; then \
                rm -rf "$dir"/dist; \
        fi \
done
find . -name '*.pyc' -exec rm -f {} +
find . -name '*.pyo' -exec rm -f {} +
find . -name '*~' -exec rm -f {} +
find . -name '__pycache__' -exec rm -fr {} +
find . -name '.DS_Store' -exec rm -fr {} +
if [ "$VIRTUAL_ENV" == "" ]];\
then\
        pipenv --rm;\
        pipenv --clear;\
        pipenv --python 3.10;\
        pipenv install --skip-lock;\
        pipenv install --dev --skip-lock;\
        pipenv run pip install -e .;\
        pipenv run pip install -e plugins/autogen;\
        pipenv run pip install -e plugins/claude;\
        pipenv run pip install -e plugins/crew_ai;\
        pipenv run pip install -e plugins/griptape;\
        pipenv run pip install -e plugins/julep;\
        pipenv run pip install -e plugins/langchain;\
        pipenv run pip install -e plugins/llamaindex;\
        pipenv run pip install -e plugins/lyzr;\
        pipenv run pip install -e plugins/openai;\
        echo "Enter virtual environment with all development dependencies now: 'pipenv shell'.";\
else\
        echo "In a virtual environment! Exit first: 'exit'.";\
fi
/bin/sh: 1: [: unexpected operator
In a virtual environment! Exit first: 'exit'.

Enhance Research Assistant
Enhance News summary

Describe the solution you'd like
Add a new local action to list files & folders at a given path locally.

Make the examples beginner friendly (To understand)

• Req.tx
• .env.example
• Python Files (Modular is better)
• Readme file
• Have gpt4o everywhere. The naming should be consistent (snake_case for variables, Functions I think is pascal case Check once)

Assigneed:
@theyashwanthsai
@Prathit-tech

Will the dashboard be open source later?

Description:
I am encountering an issue when trying to execute an action using the ComposioToolSet in the composio_langchain library. The specific error I receive is: 'sessionmaker' object has no attribute 'rollback'.

Steps to Reproduce:

Configured the ComposioToolSet with the API key.
Attempted to execute the following action:

from composio_langchain import ComposioToolSet, App

composio_toolset = ComposioToolSet(api_key="my-composio-key")
tools = composio_toolset.get_tools(apps=[App.SCHEDULER])

payload = {
    "webhook_url": "https://83f7-177-101-111-149.ngrok-free.app/google-notifications",
    "cron_pattern": "20 13 * * *",
    "start_date": "2024-06-27",
    "end_date": "2024-06-27",
    "webhook_payload": {
        "message": "Test"
    }
}

tools[0].run(tool_input=payload)

Expected Result:
The action should execute without errors, setting up the webhook as specified in the payload.

Actual Result:
I receive the following error:

HTTPError: {"message":"Failed with response: Error executing action with error: 'sessionmaker' object has no attribute 'rollback'"}

Environment:

Python version: 3.11.9
composio_langchain version: 0.3.15
Operating System: Windows 11 Pro

Additional Information:
I have tried updating all dependencies and checking the documentation, but the issue persists.
The same error is obtained when passing the tool to an agent.

I appreciate any help or guidance on how to resolve this issue.

We want TS SDK

Describe the solution you'd like
Add a new local action to retrieve the research papers using arxiv APIs. The ideal design should support pagination, filtering, max results ,min results

Describe the solution you'd like
Add a new local action to move a file or folder at a given path locally.

I tried the example code that I found here....

https://docs.composio.dev/guides/examples/todo-to-calendar#putting-it-all-together

This is working as expected. But is there any way to set the time to IST (Indian Standard Time)?

• Add All Files
• Consistent naming

This issue proposes the addition of a test enum to the Action class in enums.py to enhance functionality. The proposed change is as follows:

class Action(Enum):
    TEST = 'test'

This enum value 'TEST' serves as a placeholder for demonstration purposes. Please review and incorporate this change into the development branch.

add sql tool
add file tool
create an agent that queries sql db and logs it

Hi, the discord link in the README.md is invalid. Can you please update it?