| # KB: Meta Orchestrator & System Upgrades |
|
|
| This module documents the upgrade of the EbookBuilder agents from static prototypes to dynamic, Inference-powered tools, and the introduction of the **Meta Orchestrator**. |
|
|
| ## 1. The Meta Orchestrator (Meta Agent) |
| - **Function**: `meta_orchestrator` |
| - **Role**: The "Brain" of the system. It takes high-level user goals and decomposes them into specific tool calls. |
| - **Workflow**: |
| 1. Analyze user intent. |
| 2. Map intent to the 12+ specialized tools. |
| 3. Suggest a sequence of execution. |
|
|
| ## 2. Upgraded AI Capabilities |
| All specialized agents now use the **Hugging Face Inference API** (via `InferenceClient`) for real-time processing: |
| - **Llama-3 Integration**: Powers the Meta Agent, Fact-Checker, Illustrator, and Legal Guard for complex reasoning. |
| - **DistilBERT Integration**: Powers the Sentiment Analyst for high-speed text classification. |
| - **Style-Preserving Translation**: Uses state-of-the-art models for the Localization Agent. |
|
|
| ## 3. Operational Benefits |
| - **Real-Time Results**: No longer returns "mock" data; tools now provide actual AI-generated insights. |
| - **Scalability**: Tasks are offloaded to Hugging Face's global inference infrastructure. |
| - **Intelligence**: The Meta Agent ensures that complex requests (e.g., "Translate my book and generate a marketing plan for it") are handled with logical consistency. |
|
|
