Spaces:
Running
Running
Deploy Aureon OS (deploy/selfhost stack: nginx edge + fusion API x2 + console + HNC organ + redis + postgres; paper mode) (part 20)
8495bb8 verified | /** Server-side external-LLM routing shared by Aureon Supabase Edge Functions. */ | |
| export type ExternalLlmBody = Record<string, unknown>; | |
| export interface ExternalLlmRequest { | |
| primaryApiKey?: string | null; | |
| primaryBody: ExternalLlmBody; | |
| primaryUrl?: string; | |
| } | |
| /** Parse an OpenAI-compatible message that may wrap JSON in prose or fences. */ | |
| export function parseExternalLlmJson<T>(content: unknown, fallback: T): T { | |
| const text = String(content || "").trim(); | |
| if (!text) return fallback; | |
| const candidates = [text]; | |
| const fenced = text.match(/```(?:json)?\s*([\s\S]*?)```/i); | |
| if (fenced?.[1]) candidates.push(fenced[1].trim()); | |
| const objectStart = text.indexOf("{"); | |
| const objectEnd = text.lastIndexOf("}"); | |
| if (objectStart >= 0 && objectEnd > objectStart) { | |
| candidates.push(text.slice(objectStart, objectEnd + 1)); | |
| } | |
| for (const candidate of candidates) { | |
| try { | |
| return JSON.parse(candidate) as T; | |
| } catch { | |
| // Try the next compatible representation. | |
| } | |
| } | |
| return fallback; | |
| } | |
| const DEFAULT_PRIMARY_URL = "https://ai.gateway.lovable.dev/v1/chat/completions"; | |
| const DEFAULT_OLLAMA_URL = "https://ollama.com/v1/chat/completions"; | |
| function env(name: string): string { | |
| return String(Deno.env.get(name) || "").trim(); | |
| } | |
| function truthy(value: string): boolean { | |
| return ["1", "true", "yes", "on", "enabled"].includes(value.toLowerCase()); | |
| } | |
| function ollamaFallbackEnabled(): boolean { | |
| const route = env("AUREON_EXTERNAL_LLM_FALLBACK").toLowerCase(); | |
| if (["none", "off", "disabled", "false", "0"].includes(route)) return false; | |
| return route === "" || route === "ollama" || route === "ollama_cloud" || truthy(route); | |
| } | |
| function ollamaApiKey(): string { | |
| return env("OLLAMA_API_KEY") || env("AUREON_OLLAMA_API_KEY") || env("AUREON_LLM_API_KEY"); | |
| } | |
| function ollamaEndpoint(): string { | |
| const configured = env("AUREON_LLM_BASE_URL") || env("AUREON_OLLAMA_BASE_URL"); | |
| if (!configured) return DEFAULT_OLLAMA_URL; | |
| const base = configured.replace(/\/+$/, ""); | |
| if (/\/chat\/completions$/i.test(base)) return base; | |
| if (/\/v1$/i.test(base)) return `${base}/chat/completions`; | |
| return `${base}/v1/chat/completions`; | |
| } | |
| function ollamaBody(primaryBody: ExternalLlmBody): ExternalLlmBody { | |
| const effort = env("AUREON_OLLAMA_REASONING_EFFORT").toLowerCase(); | |
| const reasoningEffort = ["none", "low", "medium", "high"].includes(effort) ? effort : "none"; | |
| return { | |
| ...primaryBody, | |
| model: env("AUREON_LLM_MODEL") || env("AUREON_OLLAMA_MODEL") || env("OLLAMA_MODEL") || "kimi-k2.7-code", | |
| reasoning_effort: reasoningEffort, | |
| }; | |
| } | |
| function unavailableResponse(): Response { | |
| return new Response( | |
| JSON.stringify({ error: { message: "No external LLM provider is configured." } }), | |
| { status: 503, headers: { "Content-Type": "application/json" } }, | |
| ); | |
| } | |
| /** | |
| * Try the function's explicit gateway first, then Ollama Cloud. Response bodies | |
| * remain OpenAI-compatible, including streaming responses, so existing callers | |
| * can return or parse them unchanged. Credentials never leave server-side env. | |
| */ | |
| export async function fetchExternalLlm(request: ExternalLlmRequest): Promise<Response> { | |
| let primaryResponse: Response | null = null; | |
| const primaryKey = String(request.primaryApiKey || "").trim(); | |
| if (primaryKey) { | |
| try { | |
| primaryResponse = await fetch(request.primaryUrl || DEFAULT_PRIMARY_URL, { | |
| method: "POST", | |
| headers: { | |
| Authorization: `Bearer ${primaryKey}`, | |
| "Content-Type": "application/json", | |
| }, | |
| body: JSON.stringify(request.primaryBody), | |
| }); | |
| if (primaryResponse.ok) return primaryResponse; | |
| } catch { | |
| primaryResponse = null; | |
| } | |
| } | |
| const fallbackKey = ollamaApiKey(); | |
| if (ollamaFallbackEnabled() && fallbackKey) { | |
| try { | |
| return await fetch(ollamaEndpoint(), { | |
| method: "POST", | |
| headers: { | |
| Authorization: `Bearer ${fallbackKey}`, | |
| "Content-Type": "application/json", | |
| }, | |
| body: JSON.stringify(ollamaBody(request.primaryBody)), | |
| }); | |
| } catch { | |
| // Preserve an explicit provider's response when available; otherwise | |
| // return a secret-free service-unavailable response below. | |
| } | |
| } | |
| return primaryResponse || unavailableResponse(); | |
| } | |