socraticode/dist/services/provider-openai.js

SocratiCode — MCP server for local codebase indexing, semantic search, and code dependency graphs. All private, all local via Docker.

// SPDX-License-Identifier: AGPL-3.0-only
// Copyright (C) 2026 Giancarlo Erra - Altaire Limited
/**
 * OpenAI embedding provider.
 *
 * Uses the OpenAI Embeddings API (text-embedding-3-small by default).
 *
 * Required env:
 *   EMBEDDING_PROVIDER=openai
 *   OPENAI_API_KEY=sk-...
 *
 * Optional env:
 *   EMBEDDING_MODEL=text-embedding-3-small  (default)
 *   EMBEDDING_DIMENSIONS=1536               (default for text-embedding-3-small)
 */
import OpenAI from "openai";
import { getEmbeddingConfig } from "./embedding-config.js";
import { logger } from "./logger.js";
// ── Constants ───────────────────────────────────────────────────────────
/**
 * OpenAI embeddings API accepts up to 2048 inputs per request,
 * but we use a conservative batch size to stay within rate limits.
 */
const OPENAI_BATCH_SIZE = 512;
/** Max input tokens for OpenAI embedding models (8191 for text-embedding-3-*). */
const OPENAI_MAX_TOKENS = 8191;
/**
 * Conservative chars-per-token estimate for code.
 * OpenAI tokenizers average ~4 chars/token for English,
 * but code has more single-char tokens. Using 3.0 for safety.
 */
const CHARS_PER_TOKEN_ESTIMATE = 3.0;
// ── Client management ───────────────────────────────────────────────────
let openaiClient = null;
function getClient() {
    if (!openaiClient) {
        const apiKey = process.env.OPENAI_API_KEY;
        if (!apiKey) {
            throw new Error("OPENAI_API_KEY environment variable is required when using the OpenAI embedding provider. " +
                "Set it in your MCP config env block.");
        }
        openaiClient = new OpenAI({ apiKey });
    }
    return openaiClient;
}
/** Reset client (for testing). */
export function resetOpenAIClient() {
    openaiClient = null;
}
// ── Pre-truncation ──────────────────────────────────────────────────────
function pretruncateTexts(texts, contextLength) {
    if (contextLength <= 0)
        return texts;
    const maxChars = Math.floor(contextLength * CHARS_PER_TOKEN_ESTIMATE);
    return texts.map((t) => (t.length > maxChars ? t.substring(0, maxChars) : t));
}
// ── Provider class ──────────────────────────────────────────────────────
export class OpenAIEmbeddingProvider {
    name = "openai";
    async ensureReady() {
        // Validate API key is present
        getClient();
        // Verify connectivity with a minimal request
        try {
            const client = getClient();
            await client.models.list();
            logger.info("OpenAI embedding provider ready", {
                model: getEmbeddingConfig().embeddingModel,
            });
        }
        catch (err) {
            const message = err instanceof Error ? err.message : String(err);
            throw new Error(`OpenAI API is not reachable: ${message}`);
        }
        // Cloud providers don't pull models / start containers
        return { modelPulled: false, containerStarted: false, imagePulled: false };
    }
    async embed(texts) {
        if (texts.length === 0)
            return [];
        const config = getEmbeddingConfig();
        const client = getClient();
        const truncated = pretruncateTexts(texts, config.embeddingContextLength || OPENAI_MAX_TOKENS);
        // OpenAI supports batching natively — send all at once if within limits
        if (truncated.length <= OPENAI_BATCH_SIZE) {
            return this._embedBatch(client, truncated, config.embeddingModel, config.embeddingDimensions);
        }
        // Split into sub-batches
        const results = [];
        for (let i = 0; i < truncated.length; i += OPENAI_BATCH_SIZE) {
            const batch = truncated.slice(i, i + OPENAI_BATCH_SIZE);
            const embeddings = await this._embedBatch(client, batch, config.embeddingModel, config.embeddingDimensions);
            results.push(...embeddings);
        }
        return results;
    }
    async embedSingle(text) {
        const results = await this.embed([text]);
        if (results.length === 0) {
            throw new Error("Embedding failed: no result returned");
        }
        return results[0];
    }
    async healthCheck() {
        const config = getEmbeddingConfig();
        const lines = [];
        const icon = (ok) => (ok ? "[OK]" : "[MISSING]");
        const hasKey = !!process.env.OPENAI_API_KEY;
        lines.push(`${icon(hasKey)} OpenAI API key: ${hasKey ? "Configured" : "Missing — set OPENAI_API_KEY in your MCP config"}`);
        if (!hasKey) {
            return { available: false, modelReady: false, statusLines: lines };
        }
        try {
            const client = getClient();
            await client.models.list();
            lines.push(`${icon(true)} OpenAI API: Reachable`);
            lines.push(`${icon(true)} Embedding model: ${config.embeddingModel}`);
            return { available: true, modelReady: true, statusLines: lines };
        }
        catch (err) {
            const message = err instanceof Error ? err.message : String(err);
            lines.push(`${icon(false)} OpenAI API: ${message}`);
            return { available: false, modelReady: false, statusLines: lines };
        }
    }
    async _embedBatch(client, texts, model, dimensions) {
        const supportsDimensions = model.startsWith("text-embedding-3");
        const response = await client.embeddings.create({
            model,
            input: texts,
            ...(supportsDimensions ? { dimensions } : {}),
        });
        // OpenAI returns embeddings sorted by index, but we ensure order
        const sorted = response.data.sort((a, b) => a.index - b.index);
        return sorted.map((d) => d.embedding);
    }
}
//# sourceMappingURL=provider-openai.js.map