Releases | Open Source Science

https://github.com/pgalko/bambooai - v0.4.25

Rebuild of the groq_models.py module, now including tool calls - Added support for "openai/gpt-oss-120b" and "moonshotai/kimi-k2-instruct" from the Groq API - Added support for "gpt-5"and "gpt-5-mini" (The OpenAI verification requirement/process is annoying!)

- Python
Published by pgalko 7 months ago

https://github.com/pgalko/bambooai - v0.4.24

Date pickers for SweatStack window selection (https://github.com/pgalko/BambooAI/pull/47) * Changes default port in docker compose to 5001 because MacOS uses 5000 already

Date pickers for SweatStack date selection and better button loading message
Adds back lap metric for SweatStack data selection

@AartGoossens

- Python
Published by pgalko 8 months ago

https://github.com/pgalko/bambooai - v0.4.22

Add prompt enhancements for better dataset comprehension - Added dataset summary/describe in addition to df.head. This is required for composite datasets eg. SweatStack multi athlete. - Ontology rebuild to match SweatStack/Garmin datasets - Better management of analyst_selector context window - Some styling improvements

@AartGoossens

- Python
Published by pgalko 8 months ago

https://github.com/pgalko/bambooai - v0.4.21

Added export to PDF feature - When the analysis is complete, user now has an ability to export the summary and the visualisations as PDF - Updated the LLM_CONFIG.json with the new Gemini pro and Flash models

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.20

Added feature to generate and explore ideas - A new feature that allows generation and exploration of ideas. When the analysis is complete the user can request the variation of the original task to explore and execute the varied versions. User has an ability to set the degree of variation on scale 1-5. - Updated workflow map to indicate forking, as related to the previous point - Theme toggle moved under settings - A few bug fixes

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.19

A major refactor of Messages management, and some bug fixes Thanks @rnYulianto

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - 0.4.18

Integration with SweatStack

SweatStack is a Backend-as-a-Service (BaaS) for sport data applications. More information about SweatStack can be found at sweatstack.no.

Thanks @AartGoossens

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.17

Bug fixes
Custom code exec fix
Better way to handle text highlighting
Cleaned up OutputManager class and system messages from QA module

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.15

Adds manual Episodic memory search - This allows user to manually search for previous interactions stored in episodic memory - The feature requires vectordb to be enabled, and index initialised - Returns topk=10 results, filtered for score > 0.2

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.13

Fixed non-stopping spinner, added the new Gemini Pro model - Adds the new Gemini Pro 6-5 model - Added thinking budget - Fixes non-stopping tool call spinner - Fixes analyst selector instructions when web search is required

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.12

More styling.... God I hate this, I now have a lot more respect for the frontend developers and designers. I am pretty sure I have broken something...

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.11

Some styling improvements with backward compatibility

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.09

Format Gemini thoughts
Custom code exec bug fix

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.08

Major rebuild of QA Retrieval

This is a first step towards episodic memory functionality
Adds a seamless integration of the vector storage for the selected tasks
Links with the local favourite solutions storage.
When the solution is stored in the favourites, the intent_breakdown is vectorised and stored in Pinecone vector db together with the relevant metadata
When the new task lands the vector_db is searched for similar tasks from the past. If the match is found the metadata is retrieved and offered as examples to guide the solving of the new task.
Prompts have been adjusted to allow "smarter" integration of the retrieved similar past solutions

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.07

Added feature to generate datasets and files - BambooAI can now capture any datasets that resulted from the execution of the code, and consolidates them in the "datasets/generated" folder - User can view or download the generated datasets using Web App or directly from the directory - The feature works for both local and remote code execution - Fixed a substantial bug where default_prompts.yaml was not being included in the package - Version change to 0.4.07

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.06

Add ability to use custom prompt templates, and support for the new Anthropic Claude 4 models. - Changed prompt template format to yaml - Adds a new class PromptManager to handle prompt templates assignment - Adds a new parameter custom_prompt_file to allow users to submit path to their custom prompt templates yaml file, that will overwrite the default prompt templates - All prompt related handling now consolidated within messages package - Added support for the new Anthropic Claude 4 (Sonnet, Opus) models

- Python
Published by pgalko 9 months ago

https://github.com/pgalko/bambooai - v0.4.05

Enable use of auxiliary datasets and generic ontologies

Adds functionality to use auxiliary datasets in .csv or .parquet formats
- Added a parameter to provide paths to aux dataset
- Up to 3 aux datasets can be included at the time
- No need to load into memory upon instantiation. The datasets are puled and merged/joined during the code execution if they are relevant to the analysis task.
- Supported if running inside docker container, or using remote code executor API
Adds a new UI handling for Ontologies where they are now attached the same way as datasets
The framework now offers a support for generic type of ontologies, describing schema and content of one or more custom datasets
Removed feature to attach aux datasets via function calls
Removed feature to attach datasets from the Cloud storage services. Will need to update the logic, and will re-add in a future release
Some bug fixes

Screenshot 2025-05-17 at 2 22 33 pm

- Python
Published by pgalko 10 months ago

https://github.com/pgalko/bambooai - v0.4.04

Added visual workflow navigation feature - Added a new menu item "Workflow Map (CMD+M)" to visually inspect and navigate the branches of the current workflow. This has several benefits including ease of navigation, better context window management, better response accuracy, better management of checkpoints etc. - The feature is based on interactive mermaid diagrams tracking the flow and branching of the enquiry

Workflow Map Feature

- Python
Published by pgalko 10 months ago

https://github.com/pgalko/bambooai - v0.4.03

Improvements, bug fixes
Added support for the new Gemini 2.5 Pro model

- Python
Published by pgalko 10 months ago

https://github.com/pgalko/bambooai - v0.4.02

bug fixes

- Python
Published by pgalko 10 months ago

https://github.com/pgalko/bambooai - v0.4.01

Fixed import errors Added support for Docker

- Python
Published by pgalko 10 months ago

https://github.com/pgalko/bambooai - v0.4.00

This is a major upgrade

A new web app (Demo in README.md)
Bunch of new features and bug fixes
Tighter integration of reasoning models
Incorporated user feedback during execution
Added support for VLLM and Ollama
Custom code edits
Workflows storage
Multimodal and In-Context queries

- Python
Published by pgalko 10 months ago

https://github.com/pgalko/bambooai - v0.3.52

This is a relatively large, and potentially breaking update.

Added support for Anthropic Claude 3.5 including function calls (streaming)
Major overhaul to vector storage and retrieval
Added a functionality when data frames can be explained/described with ontology. This significantly improves the accuracy of the responses.
Added a new agent "Dataframe Inspector"
A bunch of changes to prompts to facilitate the new features

- Python
Published by pgalko over 1 year ago

https://github.com/pgalko/bambooai - v0.3.50

Library now supports scraping of dynamic web content via Selenium
Requires manual ChromeDriver download, and the path to it set in SELENIUM_WEBDRIVER_PATH env var
If the env var is set, the library selects the Selenium for all scraping tasks
a couple bug fixes

- Python
Published by pgalko almost 2 years ago

https://github.com/pgalko/bambooai - v0.3.48

Major refactor of qa_retrieval.py, new gemini models

Adds support for new pinecone client
Removes vector db records duplication
Add support for OpenAI embeddings models in addition to hf_sentence_transformers
text-embedding-3-small now default embeddings model

- Python
Published by pgalko almost 2 years ago

https://github.com/pgalko/bambooai - v0.3.44

Google search seamlessly incorporated into the flow. A very positive results with Groq:Llama 3 70B when selected as a model for the Search agent.

json {"agent": "Google Search Query Generator", "details": {"model": "llama3-70b-8192", "provider":"groq","max_tokens": 4000, "temperature": 0}}, {"agent": "Google Search Summarizer", "details": {"model": "llama3-70b-8192", "provider":"groq","max_tokens": 4000, "temperature": 0}} - Some improvements to jupyter notebook output formatting. - Search could benefit further from Something like Selenium or pypeteer to allow for scraping of dynamic websites. At the moment only static content is supported. Tricky, as we do not want the library to become too bloated

- Python
Published by pgalko almost 2 years ago

https://github.com/pgalko/bambooai - v0.3.42

I have now updated the notebook output formatting using Markdown instead of HTML. It is now much more pleasant for the user. The changes are included in the this version (v0.3.42), and pushed to PyPi.
Video to illustrate the new output here: https://github.com/pgalko/BambooAI/assets/39939157/6058a3a2-63d9-44b9-b065-0a0cda5d7e17
Also benchmarked the library against "OpenAI Assitants API + Code Interpreter". BambooAI much cheaper and faster :-).

Task: Devise a machine learning model to predict the survival of passengers on the Titanic. The output should include the accuracy of the model and visualizations of the confusion matrix, correlation matrix, and other relevant metrics.

Dataset: Titanic.csv

Model: GPT-4-Turbo

OpenAI Assistants API (Code Interpreter)

Result:
- Confusion Matrix:
- True Negative (TN): 90 passengers were correctly predicted as not surviving.
- True Positive (TP): 56 passengers were correctly predicted as surviving.
- False Negative (FN): 18 passengers were incorrectly predicted as not surviving.
- False Positive (FP): 15 passengers were incorrectly predicted as surviving.

| Metric | Value | | -------------- | ------------ | | Execution Time | 77.12 seconds| | Input Tokens | 7128 | | Output Tokens | 1215 | | Total Cost | $0.1077 |

BambooAI (No Planning, Google Search or Vector DB)

Result:
- Confusion Matrix:
- True Negative (TN): 92 passengers were correctly predicted as not surviving.
- True Positive (TP): 55 passengers were correctly predicted as surviving.
- False Negative (FN): 19 passengers were incorrectly predicted as not surviving.
- False Positive (FP): 13 passengers were incorrectly predicted as surviving.

| Metric | Value | | -------------- | ------------ | | Execution Time | 47.39 seconds| | Input Tokens | 722 | | Output Tokens | 931 | | Total Cost | $0.0353 |

- Python
Published by pgalko almost 2 years ago

https://github.com/pgalko/bambooai - v0.3.38

Planning agent can now use google search via function_calls. This is currently only available for OpenAI LLMs.
A new logic for expert selector
Plans now included in Vector DB record metadata alongside code. This is particularly beneficial for non OpenAI models.
A completely new google_search.py module using ReAct method
Some prompt adjustments. Current date now included in some system prompts.
A bunch of bug fixes

- Python
Published by pgalko almost 2 years ago

https://github.com/pgalko/bambooai - v0.3.32

Added support for a bunch of models and APIs

Added support for Ollama
Added support for Anthtropic, Mistral, Groq, Google Gemini
Some bug fixes

- Python
Published by pgalko almost 2 years ago

https://github.com/pgalko/bambooai - v0.3.30

BambooAI now compatible with the latest version of OpenAI client library
New OpenSource LLM added "Open Code Interpreter"
A few mods to prompts
The Default OpenAI model switched to GPT-4-Turbo

- Python
Published by pgalko almost 2 years ago

https://github.com/pgalko/bambooai - v0.3.29

Load llm config from env var or json file
Load prompt templates from json file
Add ability to specify an llm config individually for each agent
Append full traceback to error correction calls
Refactor the code for functions and classes to match agent work flow
Change variable names to be more descriptive
Change output messages to be more descriptive

Deprecation Notice (October 25, 2023): Please note that the "llm", "localcodemodel", "llmswitchplan", and "llmswitchcode" parameters have been deprecated as of v 0.3.29. The assignment of models and model parameters to agents is now handled via LLMCONFIG. This can be set either as an environment variable or via a LLMCONFIG.json file in the working directory. Please see README "Usage" section for details.