Files
memex/notes/org-skill-embedding-generator.org

81 lines
2.7 KiB
Org Mode

:PROPERTIES:
:ID: 6cd0b4d7-e7cf-49ba-a847-e81c4ace0f76
:CREATED: [2026-04-04 Sat 20:27]
:EDITED: [2026-04-07 Tue 13:42]
:END:
#+TITLE: SKILL: Embedding Generator (Universal Literate Note)
#+STARTUP: content
#+FILETAGS: :embeddings:neural:psf:
* Overview
The *Embedding Generator* provides vector representations of text for semantic search and neural memory.
* Phase A: Demand (PRD)
:PROPERTIES:
:STATUS: FROZEN
:END:
** 1. Purpose
Generate embeddings for text strings.
** 2. User Needs
- *Provider Choice:* Support for local (Ollama) or remote (Gemini, OpenAI) providers.
- *Batching:* Efficiency through batching text (future-proof).
* Phase B: Blueprint (PROTOCOL)
:PROPERTIES:
:STATUS: SIGNED
:END:
** Phase B: Blueprint (PROTOCOL)
:PROPERTIES:
:STATUS: DRAFT
:END:
*** 1. Architectural Intent
This system aims for a flexible and extensible architecture that can accommodate different embedding providers (local or remote) while maintaining a consistent interface for the user. It also seeks to optimize for batch processing to improve throughput. The core design principle is *provider abstraction*.
*** 2. Semantic Interfaces
#+BEGIN_SRC lisp
;;; Primary Function: Generate Embeddings
;;; Input: A list of strings, and provider configuration
;;; Output: A list of embedding vectors (lists of floats).
(defun generate-embeddings (texts provider-config)
"""Generates embeddings for a list of texts using the specified provider.""")
;;; Provider Configuration Structure
;;; This plist defines the provider to use, and any necessary credentials or parameters.
;;; Example: Local Ollama provider
;;; (:provider :ollama :model "mistralai/Mistral-7B-Instruct-v0.2")
;;; Example: Remote Gemini provider
;;; (:provider :gemini :api-key "YOUR_API_KEY" :model "gemini-1.5-pro")
;;; Example: Remote OpenAI provider.
;;; (:provider :openai :api-key "YOUR_API_KEY" :model "text-embedding-ada-002")
;;; Sub-Function: (Abstract) Provider-Specific Embedding Generation
;;; This function is implemented differently for each provider.
(defgeneric generate-embeddings-from-provider (texts provider-config))
;;; Example implementation for :ollama provider
(defmethod generate-embeddings-from-provider (texts (provider-config (eql (getf provider-config :provider :ollama))))
"""Generates embeddings using a local Ollama server.""")
;;; Example implementation for :gemini provider
(defmethod generate-embeddings-from-provider (texts (provider-config (eql (getf provider-config :provider :gemini))))
"""Generates embeddings using the Gemini API.""")
;;; Example implementation for :openai provider
(defmethod generate-embeddings-from-provider (texts (provider-config (eql (getf provider-config :provider :openai))))
"""Generates embeddings using the OpenAI API.""")
#+END_SRC