Objective: Figure out the workflow and potential limits/road blocks for the functionality:
<Send User Prompt to GEMINI, return the React Flow tree data in JSON, store in Database>
Backend send user prompts to GEMINI API [hardcode user input/prompt - text only] - 4 hours
Receive and store the returned JSON under the same user + input to Mongo Database, [Please refer to Schemas] - Please contact @ChTiSh for testing database credentials - 2 hours
Acceptance Criteria:
AC1: API Response Handling
Given the GEMINI API receives the user input,
When GEMINI processes the input and generates tree data,
Then the API should send this data back to the frontend in JSON format.
AC2: Data Storage
Given the frontend receives JSON tree data,
When the data is received,
Then it should be stored in MongoDB according to the specified schema.
AC3: Explore in Sizes and Time (Gemini 1.0 and Gemini 1.0 Pro) - 1.5 has too litttle limit
Explore 3 different user input size (30 characters, 3000 characters, 30000 characters)
Record Time and Token Used for each type
Stretch Goal:
Can gemini generate react-flow code instead of just JSON for nodes and edges?
Objective: Figure out the workflow and potential limits/road blocks for the functionality: <Send User Prompt to GEMINI, return the React Flow tree data in JSON, store in Database>
PreReq: Initial Study for prompting, see Gemini cookbook:https://github.com/google-gemini/cookbook - 4 hours Gemini API Key + Project Setup for Testing - 1 hour
Please refer to Darren's gemini PR: https://github.com/Linkta-org/core/pull/64
Ticket Scope:
Acceptance Criteria: AC1: API Response Handling
AC2: Data Storage
AC3: Explore in Sizes and Time (Gemini 1.0 and Gemini 1.0 Pro) - 1.5 has too litttle limit
Stretch Goal: Can gemini generate react-flow code instead of just JSON for nodes and edges?
Reference: For gemini-1.0-pro, only the text field is valid. The token limit is 32,760. For gemini-1.0-pro-vision, you may specify either text only, text and up to 16 images, or text and 1 video clip. The token limit is 16,384. Note: Gemini updated their limit on their website to significantly lower the 1.0 from 60 calls to 15 calls a couple days ago. Gemini limit: https://ai.google.dev/gemini-api/docs/models/gemini#rate-limits https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/gemini#:~:text=The%20token%20limit%20is%2032%2C760,%2C%20video%2C%20and%20audio%20files. https://cloud.google.com/vertex-ai?hl=en