evOo-local/packages/adapters/src/media-context.ts

import type {
  AgentInputImage,
  AgentRunModel,
  AgentRuntime,
  ArtifactStore,
  JobPublisher,
  LinkShortener,
  MediaInferenceProvider,
  MediaPublisher,
  MediaRenderer,
  MediaSourceProvider,
  MotionGraphicsRenderer,
} from "@rakazo/adapter-kit";
import type { ModelTier } from "@rakazo/contracts";
import type { MediaRepos, PrismaClient } from "@rakazo/db";
import { getLogger } from "@rakazo/logging";
import type { ProfileSet } from "@rakazo/media";
import type { z } from "zod";
import { aiRecipient } from "./ai-consent.js";

/*
 * Shared wiring for Media Studio services. Composition roots (API, worker)
 * build one MediaServiceDeps; every media service, job handler, and agent tool
 * receives it instead of reaching for globals.
 */

/** Realtime notifications the web and mobile clients subscribe to per space. */
export type MediaRealtimeEvent =
  | { type: "media.source"; sourceId: string; status: string; error?: string | null }
  | { type: "media.asset"; assetId: string; ingestState: string }
  | { type: "media.collections"; kind: "session" | "theme" }
  | { type: "media.people" }
  | { type: "media.render"; renderId: string; projectId: string; status: string; progress: number }
  | { type: "media.publication"; publicationId: string; status: string };

/**
 * Structured model calls on behalf of a user in a space. Returns null instead of
 * calling a model when no model is configured, the runtime is scripted, or the
 * user has not consented to sharing data with the resolved model recipient.
 * Callers must then fall back to deterministic behavior.
 */
export interface MediaModelClient {
  generate<T>(input: {
    spaceId: string;
    userId: string;
    /** Short system instructions; the reply must be one JSON document. */
    instructions: string;
    prompt: string;
    images?: AgentInputImage[];
    schema: z.ZodType<T>;
    tier?: ModelTier;
    signal?: AbortSignal;
  }): Promise<T | null>;
}

export interface MediaServiceDeps {
  prisma: PrismaClient;
  repos: MediaRepos;
  jobs: JobPublisher;
  /** Null when no sidecar is configured; media then stays hidden in the UI. */
  inference: MediaInferenceProvider | null;
  renderer: MediaRenderer;
  motion: MotionGraphicsRenderer;
  publisher: MediaPublisher;
  shortener: LinkShortener | null;
  sources: Partial<Record<"folder" | "emulator", MediaSourceProvider>>;
  /** Absolute DATA_DIR/media. Every stored media path is relative to it. */
  mediaRoot: string;
  /** Builtin plus private platform profiles, cached by the composition root. */
  profiles: () => Promise<ProfileSet>;
  artifacts: ArtifactStore;
  model: MediaModelClient;
  publish: (spaceId: string, event: MediaRealtimeEvent) => Promise<void>;
  now: () => Date;
}

/** Realtime topic for a space's media events. */
export function mediaTopic(spaceId: string): string {
  return `media:${spaceId}`;
}

const MODEL_TIMEOUT_MS = 120_000;
const RUNTIME_FAILURE = /^(?:I hit a problem:|Unknown model )/i;

/** Extract the first JSON object or array from a model reply (models may wrap or think aloud). */
export function extractJson(text: string): unknown {
  const trimmed = text.trim();
  const fenced = /```(?:json)?\s*([\s\S]*?)```/i.exec(trimmed);
  const body = fenced?.[1] ?? trimmed;
  const start = body.search(/[[{]/);
  if (start < 0) return undefined;
  const open = body[start];
  const close = open === "{" ? "}" : "]";
  const end = body.lastIndexOf(close);
  if (end <= start) return undefined;
  try {
    return JSON.parse(body.slice(start, end + 1));
  } catch {
    return undefined;
  }
}

export function createMediaModelClient(deps: {
  prisma: PrismaClient;
  runtime: AgentRuntime;
  resolveModel: (scope: {
    userId: string;
    spaceId: string;
    tier?: ModelTier;
  }) => Promise<AgentRunModel>;
}): MediaModelClient {
  return {
    async generate(input) {
      if (!deps.runtime.describe().capabilities.compaction) return null;
      let model: AgentRunModel;
      try {
        model = await deps.resolveModel({
          userId: input.userId,
          spaceId: input.spaceId,
          tier: input.tier ?? "fast",
        });
      } catch {
        return null;
      }
      if (model.provider === "scripted") return null;
      const recipient = aiRecipient({
        provider: model.provider,
        use: "model",
        modelId: model.id,
        baseUrl: model.baseUrl,
      });
      if (!recipient) return null;
      const consent = await deps.prisma.aiDataConsent.findUnique({
        where: {
          userId_spaceId_recipientKey: {
            userId: input.userId,
            spaceId: input.spaceId,
            recipientKey: recipient.key,
          },
        },
      });
      if (!consent) return null;
      const images =
        input.images && model.acceptsImages !== false
          ? input.images.slice(0, model.maxImagesPerPrompt ?? input.images.length)
          : undefined;
      const runId = `media-${crypto.randomUUID()}`;
      const signal = input.signal
        ? AbortSignal.any([input.signal, AbortSignal.timeout(MODEL_TIMEOUT_MS)])
        : AbortSignal.timeout(MODEL_TIMEOUT_MS);
      let reply = "";
      try {
        for await (const event of deps.runtime.run(
          {
            botId: "media-studio",
            threadId: runId,
            runId,
            prompt: input.prompt,
            instructions: input.instructions,
            history: [],
            tools: [],
            model,
            ...(images && images.length > 0 ? { currentTurnImages: images } : {}),
          },
          {
            operationId: runId,
            traceId: runId,
            spaceId: input.spaceId,
            userId: input.userId,
            signal,
          },
        )) {
          if (event.type === "done" && event.text) reply = event.text.trim();
        }
      } catch (error) {
        getLogger().warn("media model call failed", {
          reason: error instanceof Error ? error.message : String(error),
        });
        return null;
      }
      if (!reply || RUNTIME_FAILURE.test(reply)) return null;
      const parsed = input.schema.safeParse(extractJson(reply));
      return parsed.success ? parsed.data : null;
    },
  };
}