import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";

/**
 * New API gateway (https://github.com/Calcium-Ion/new-api) aggregation provider.
 *
 * Routes a self-hosted New API instance — which normalizes many upstream
 * vendors (Zhipu GLM, Kimi/Moonshot, Anthropic, OpenAI, …) behind a single
 * OpenAI-compatible `/v1/chat/completions` endpoint — as one Pi provider.
 *
 * Configure with:
 *   NEW_API_KEY=sk-...          # token issued by the New API dashboard
 *   NEW_API_BASE_URL=http://…   # optional override (defaults to the address below)
 *
 * Model `compat` blocks mirror the built-in pi-ai provider definitions so the
 * gateway behaves identically to talking to each vendor directly:
 *   - GLM 5.2 / 4.7  ← zai-coding-cn (thinkingFormat "zai", zaiToolStream)
 *   - K3 / K3-256K   ← moonshotai kimi-k3 (thinkingFormat "openai", deferred
 *                      tools, reasoning_effort, reasoning_content replay)
 *   - kimi-for-coding← moonshotai kimi-k2.7-code (thinkingFormat "deepseek")
 */
export default function (pi: ExtensionAPI) {
  // Gateway address. Resolved once, when the extension loads, from the
  // environment so it can be overridden without touching code. `apiKey` below
  // still uses pi's `$ENV` config-value syntax so pi can cache/refresh it.
  const baseUrl =
    process.env.NEW_API_BASE_URL || "http://10.30.60.169:3000/v1";

  pi.registerProvider("new-api", {
    name: "New API Gateway",
    baseUrl,
    apiKey: "$NEW_API_KEY",
    authHeader: true, // New API tokens are sent as `Authorization: Bearer <key>`
    api: "openai-completions",
    models: [
      // ── 智谱 GLM 5.2 ──────────────────────────────────────
      // Mirrors pi-ai `zai-coding-cn` glm-5.2 definition.
      {
        id: "glm-5.2",
        name: "GLM 5.2",
        reasoning: true,
        thinkingLevelMap: {
          minimal: null,
          low: "high",
          medium: "high",
          high: "high",
          max: "max",
        },
        input: ["text"],
        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
        contextWindow: 1000000,
        maxTokens: 131072,
        compat: {
          supportsStore: false,
          supportsDeveloperRole: false,
          supportsReasoningEffort: true,
        },
      },
      // ── 智谱 GLM 4.7 ──────────────────────────────────────
      // Mirrors pi-ai `zai-coding-cn` glm-4.7 definition.
      {
        id: "glm-4.7",
        name: "GLM 4.7",
        reasoning: true,
        input: ["text"],
        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
        contextWindow: 204800,
        maxTokens: 131072,
        compat: {
          supportsStore: false,
          supportsDeveloperRole: false,
          supportsReasoningEffort: false,
        },
      },
      // ── Kimi K3 ───────────────────────────────────────────
      // Mirrors pi-ai `moonshotai` / `moonshotai-cn` kimi-k3 definition,
      // exposed through the gateway under the official id `k3`.
      {
        id: "k3",
        name: "Kimi K3",
        reasoning: true,
        input: ["text", "image"],
        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
        contextWindow: 1048576,
        maxTokens: 131072,
        thinkingLevelMap: {
          off: null,
          minimal: null,
          low: "low",
          medium: null,
          high: "high",
          xhigh: null,
          max: "max",
        },
        compat: {
          supportsStore: false,
          supportsDeveloperRole: false,
          supportsReasoningEffort: true,
          maxTokensField: "max_tokens",
          supportsStrictMode: false,
          thinkingFormat: "openai",
          requiresReasoningContentOnAssistantMessages: true,
          deferredToolsMode: "kimi",
        },
      },
      // ── Kimi K3 (256K context) ────────────────────────────
      // Same upstream model as `k3`; the gateway offers a shorter context
      // variant. Only the context window differs.
      {
        id: "k3-256k",
        name: "Kimi K3 256K",
        reasoning: true,
        input: ["text", "image"],
        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
        contextWindow: 262144,
        maxTokens: 131072,
        thinkingLevelMap: {
          off: null,
          minimal: null,
          low: "low",
          medium: null,
          high: "high",
          xhigh: null,
          max: "max",
        },
        compat: {
          supportsStore: false,
          supportsDeveloperRole: false,
          supportsReasoningEffort: true,
          maxTokensField: "max_tokens",
          supportsStrictMode: false,
          thinkingFormat: "openai",
          requiresReasoningContentOnAssistantMessages: true,
          deferredToolsMode: "kimi",
        },
      },
      // ── Kimi for Coding (K2.7 Code) ───────────────────────
      // Mirrors pi-ai `moonshotai` / `moonshotai-cn` kimi-k2.7-code definition,
      // exposed through the gateway under the official id `kimi-for-coding`.
      {
        id: "kimi-for-coding",
        name: "Kimi For Coding",
        reasoning: true,
        input: ["text", "image"],
        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
        contextWindow: 262144,
        maxTokens: 262144,
        thinkingLevelMap: {
          off: null,
        },
        compat: {
          supportsStore: false,
          supportsDeveloperRole: false,
          supportsReasoningEffort: false,
          maxTokensField: "max_tokens",
          supportsStrictMode: false,
          thinkingFormat: "deepseek",
        },
      },
    ],
  });
}