/**
 * Shared fixtures for TASK-rectification-grounding-20260927.
 *
 * `rectification-grounding-aa-30min.golden.json` is a real local engine
 * response (`runV9CandidateScore` request + response) for a public AA chart
 * (Angelina Jolie, 1975-06-04 Los Angeles, 08:54–09:24 search window, three
 * public dated events). No private data.
 */
import { readFileSync } from "node:fs";
import {
  CASE_ID,
  FOCUS_ID,
  RESULT_ID,
  TURN_ID,
  computeFixture,
  dossierFixture,
} from "./rectification-v9-test-support.ts";
import { runV9CandidateScore } from "../src/lib/rectification-agentic/v9/engine-client.ts";
import { buildCaseInferenceState } from "../src/lib/rectification-agentic/v9/inference-adapter.ts";
import { requestCandidateIntervals, candidatePositionFields } from "../src/lib/rectification-agentic/core/candidate-window.ts";
import { candidateRangeFingerprint, evidenceLedgerFingerprint, parseV9CaseDossier } from "../src/lib/rectification-agentic/v9/tool-service.ts";

export const AA_GOLDEN = JSON.parse(readFileSync(
  new URL("./fixtures/rectification-grounding-aa-30min.golden.json", import.meta.url),
  "utf8",
)) as { request: Record<string, unknown> & { events: Array<Record<string, string>> }; response: Record<string, unknown> };

const req = AA_GOLDEN.request as Record<string, unknown> & { events: Array<Record<string, string>> };

export const AA_BASELINE = {
  birth_date: req.birth_date,
  latitude: req.lat,
  longitude: req.lon,
  timezone_offset: req.tz,
  timezone_id: "America/Los_Angeles",
  birth_time_source: "family_vague",
  reported_birth_time: "09:09",
  active_birth_time: null,
  uncertainty_before_minutes: 15,
  uncertainty_after_minutes: 15,
  birth_place_label: "Los Angeles",
};

const baseRange = { start_time: String(req.start_time), end_time: String(req.end_time) };
export const AA_RANGE = {
  ...baseRange,
  candidate_intervals: requestCandidateIntervals(AA_BASELINE as never, baseRange as never),
};

const VEDASTRO_PASS = {
  status: "passed",
  can_confirm_exact_minute: false,
  event_validation: { search_events_primary_supports_local_winner: true },
  minute_sensitive_validation: { status: "passed" },
};

/** Serve the golden engine response (and a passing VedAstro check) to every fetch. */
export function stubGoldenFetch(): () => void {
  const previous = globalThis.fetch;
  globalThis.fetch = (async (url: unknown) => {
    const body = String(url).includes("vedastro") ? VEDASTRO_PASS : AA_GOLDEN.response;
    return new Response(JSON.stringify(body), { status: 200 });
  }) as typeof fetch;
  return () => {
    globalThis.fetch = previous;
  };
}

export const AA_EVIDENCE_ROWS = req.events.map((event, index) => ({
  id: `44444444-4444-4444-8444-44444444444${index}`,
  source_turn_id: TURN_ID,
  subject: "self",
  event_kind: event.event_kind,
  domain: event.domain,
  occurred_from: event.date_start,
  occurred_to: null,
  date_precision: event.precision,
  summary: event.summary,
  status: "confirmed",
  supersedes_evidence_id: null,
  created_at: "2026-08-12T10:00:06.000Z",
}));

export const AA_TURNS = [
  {
    id: TURN_ID,
    role: "user",
    text: "2000 年拿了一个大奖；2014 年 8 月结婚。",
    status: "completed",
    created_at: "2026-08-12T10:00:00.000Z",
    completed_at: "2026-08-12T10:00:05.000Z",
  },
  {
    id: "77777777-7777-4777-8777-777777777771",
    role: "assistant",
    text: "记下了：2000 年获奖、2014 年 8 月结婚。",
    status: "completed",
    created_at: "2026-08-12T10:00:06.000Z",
    completed_at: "2026-08-12T10:00:07.000Z",
  },
];

/** The stored snapshot a real run of the golden case persists (9 candidates). */
export async function buildGoldenLatest() {
  const restore = stubGoldenFetch();
  try {
    process.env.RECTIFICATION_ALGORITHM_VERSION = String(AA_GOLDEN.response.algorithm_version);
    process.env.RECTIFICATION_DECISION_POLICY_VERSION = String(AA_GOLDEN.response.decision_policy_version);
    const scored = await runV9CandidateScore({
      baselineBirthSnapshot: AA_BASELINE,
      candidateRange: AA_RANGE as never,
      events: req.events,
    } as never);
    const parsedEvidence = parseV9CaseDossier(dossierFixture({ evidence: AA_EVIDENCE_ROWS }))!.evidence;
    const inference = buildCaseInferenceState({
      range: baseRange,
      candidates: scored.candidates as never,
      evidence: parsedEvidence as never,
      probes: [],
      transitions: scored.windowScan?.transitions as never,
    });
    const compute = { ...computeFixture({ baselineBirthSnapshot: AA_BASELINE as never }), candidate_range: AA_RANGE };
    const latest = {
      result_id: RESULT_ID,
      candidates: scored.candidates.map((candidate) => ({
        ...candidatePositionFields(candidate as never),
        candidate_id: candidate.candidateId,
        time: candidate.time,
        rank: candidate.rank,
        relative_support: candidate.relativeSupport,
        tied_minute_count: candidate.tiedMinuteCount,
        ...(candidate.clusterTimes ? { cluster_times: candidate.clusterTimes } : {}),
        ...(candidate.clusterStart ? { cluster_start: candidate.clusterStart } : {}),
        ...(candidate.clusterEnd ? { cluster_end: candidate.clusterEnd } : {}),
      })),
      overall_confidence: scored.overallConfidence,
      selection_allowed: scored.selectionAllowed,
      confirmation_allowed: false,
      decision_receipt: { ...scored.decisionReceipt, inference_state: inference },
      execution_ledger: scored.executionLedger,
      representative_time: scored.representativeTime,
      selected_time: null,
      selection_kind: null,
      evidence_ledger_fingerprint: evidenceLedgerFingerprint(parsedEvidence),
      candidate_range_fingerprint: candidateRangeFingerprint(AA_RANGE as never, compute.baseline_profile_fingerprint),
      skill_version: "9.0.0",
      algorithm_version: scored.algorithmVersion,
      event_contract_version: scored.eventContractVersion,
      decision_policy_version: scored.policyVersion,
      created_at: "2026-08-12T10:05:00.000Z",
      invalidated_at: null,
    };
    return { latest, compute, scored };
  } finally {
    restore();
  }
}

export function setFocusHandler(_fn: string, args: Record<string, unknown>) {
  return {
    focus: {
      id: FOCUS_ID,
      case_id: CASE_ID,
      question_id: args.p_question_id,
      intent: args.p_intent,
      target_evidence_id: args.p_target_evidence_id,
      target_domain: args.p_target_domain,
      target_kind: args.p_target_kind,
      expected_answer_schema: args.p_expected_answer_schema,
      status: "active",
      asked_at: "2026-08-27T00:00:00.000Z",
      resolved_at: null,
      asked_turn_id: args.p_asked_turn_id ?? null,
    },
    idempotent: false,
  };
}

export type ScriptPart = { text?: string; tool?: { name: string; input: Record<string, unknown> } };
export type ScriptTurn = { parts: ScriptPart[]; finish: string };
export type RecordedCall = { prompt: Array<{ role: string; content: unknown }>; tools: unknown };

/** An AI SDK v2 language model that replays a script and records every prompt it is sent. */
export function scriptedModel(script: ScriptTurn[], onCall?: (callNumber: number) => void) {
  const calls: RecordedCall[] = [];
  let call = 0;
  const next = (options: { prompt: unknown[]; tools?: unknown }) => {
    calls.push({ prompt: options.prompt as RecordedCall["prompt"], tools: options.tools });
    const turn = script[call] ?? { parts: [{ text: "" }], finish: "stop" };
    call += 1;
    onCall?.(call);
    return turn;
  };
  const model = {
    specificationVersion: "v2",
    provider: "fake",
    modelId: "fake-rectification",
    supportedUrls: {},
    async doGenerate(options: { prompt: unknown[]; tools?: unknown }) {
      const turn = next(options);
      const content = turn.parts.map((part, index) => part.tool
        ? { type: "tool-call", toolCallId: `g${call}-${index}`, toolName: part.tool.name, input: JSON.stringify(part.tool.input) }
        : { type: "text", text: part.text ?? "" });
      return { content, finishReason: turn.finish, usage: { inputTokens: 10, outputTokens: 10, totalTokens: 20 }, warnings: [] };
    },
    async doStream(options: { prompt: unknown[]; tools?: unknown }) {
      const turn = next(options);
      const id = `t${call}`;
      const stream = new ReadableStream({
        start(controller) {
          controller.enqueue({ type: "stream-start", warnings: [] });
          let open = false;
          for (const [index, part] of turn.parts.entries()) {
            if (part.text !== undefined) {
              if (!open) {
                controller.enqueue({ type: "text-start", id });
                open = true;
              }
              controller.enqueue({ type: "text-delta", id, delta: part.text });
            }
            if (part.tool) {
              if (open) {
                controller.enqueue({ type: "text-end", id });
                open = false;
              }
              controller.enqueue({
                type: "tool-call",
                toolCallId: `c${call}-${index}`,
                toolName: part.tool.name,
                input: JSON.stringify(part.tool.input),
              });
            }
          }
          if (open) controller.enqueue({ type: "text-end", id });
          controller.enqueue({ type: "finish", finishReason: turn.finish, usage: { inputTokens: 10, outputTokens: 10, totalTokens: 20 } });
          controller.close();
        },
      });
      return { stream };
    },
  };
  return { model, calls };
}

/** The text a client ends up showing after applying `answer.delta` events in order. */
export function clientStates(events: ReadonlyArray<Record<string, unknown>>): string[] {
  let text = "";
  const states: string[] = [];
  for (const event of events) {
    if (event.type !== "answer.delta") continue;
    const delta = String(event.text ?? "");
    text = event.replace === true ? delta : `${text}${delta}`;
    states.push(text);
  }
  return states;
}

const CJK = /[⺀-鿿豈-﫿　-〿＀-￯]/g;

/**
 * Token estimate used for the before/after size tables: one token per CJK
 * character (including full-width punctuation) plus one token per four other
 * characters. It is an estimate, not a provider tokenizer.
 */
export function estimateTokens(text: string): number {
  const cjk = (text.match(CJK) ?? []).length;
  return Math.round(cjk + (text.length - cjk) / 4);
}

export function contentText(content: unknown): string {
  return typeof content === "string" ? content : JSON.stringify(content);
}