steipete/oracle
Oracle
Oracle is an LLM API gateway client and prompt orchestrator designed to route requests between official API endpoints and automated web browser channels. It functions as a context manager and session logger, bundling local files and prompts into requests while maintaining a history of interactions.
The system distinguishes itself through a request router that can redirect traffic to alternate endpoints, such as self-hosted gateways or proxy services. It includes a token estimator that calculates request size and predicts costs using local summaries without making network calls.
The project provides capabilities for AI context management via glob-based file filtering and preference management for storing engine, model, and prompt settings. It also features session history management to list, replay, or prune past interactions from local storage.
Features
- Large Language Model Connectors - Connects local data and custom settings to LLM APIs through configurable gateways and proxy services.
- File-Based Context Selection - Filters local files using glob patterns to bundle relevant document context for LLM requests.
- LLM Context Compilers - Filters local files via glob patterns to provide specific document context to large language models.
- LLM Gateways - Acts as a request router that redirects traffic between official API endpoints and automated browser channels.
- Prompt Engineering - Combines local files and custom prompts to provide language models with specific context for complex tasks.
- Prompt Orchestration - Bundles local files and prompts into requests for LLMs with support for custom API endpoints.
- Prompt File Packers - Bundles local project files into AI prompts using pattern-based inclusion and exclusion to optimize context.
- Request Redirection - Implements request redirection to switch between official API endpoints and browser-based automation.
- API Endpoint Routing - Provides the capability to redirect LLM API requests to custom server addresses or proxy services.
- Prompt Payloads - Combines user text and local file contents into a single payload for large language model processing.
- API and Browser Channel Routing - Allows switching between official APIs and automated browser channels to access different account levels.
- Session History Retrieval - Provides a system to record and retrieve historical interaction data to restore or review session context.
- LLM Token Counters - Calculates total request size and predicts costs for LLM interactions using local token counting.
- Input Token Estimators - Calculates potential request size and predicts costs using a local token estimation mechanism.
- Glob Pattern Selectors - Uses glob patterns and exclusion rules to filter which local files are included in an AI prompt.
- Engine Routing Configurations - Directs requests to different API endpoints or browser channels based on a local persistent settings file.
- Replayable Session Logs - Records every request and response to a local storage system for later retrieval and replay.
- Pre-Request Cost Estimation - Calculates total token usage for bundled prompts to predict API costs before sending requests.
Star history
How this analysis was created: This summary and feature list were written by an AI model that read the project's README and public documentation pages. Each feature links to the documentation it came from; stars, license and language come straight from the GitHub API. The model does not read the source code, and the analysis is refreshed when the project is re-analysed. Learn more on our About page.
