@fudrouter/fsrouter 0.6.64 → 0.6.65

package/open-sse/services/combo.js

@@ -0,0 +1,571 @@
+/**
+ * Shared combo (model combo) handling with fallback support
+ */
+
+import { checkFallbackError, formatRetryAfter } from "./accountFallback.js";
+import { unavailableResponse } from "../utils/error.js";
+import { getCapabilitiesForModel } from "../providers/capabilities.js";
+import { extractTextContent } from "../translator/formats/gemini.js";
+
+// Hard capabilities = input modalities; missing one drops request data (e.g. image
+// stripped). Must be prioritized. Soft (e.g. search) only degrades a feature.
+const HARD_CAPS = new Set(["vision", "pdf", "audioInput", "videoInput"]);
+
+// Prefixes used when flattening tool turns into plain prose for panel models.
+const TOOL_CALL_PREFIX = "[Called tools: ";
+const TOOL_RESULT_PREFIX = "[Tool result: ";
+
+// Flatten tool turns into prose so panel models keep the context but can't loop
+// on tools: drop the request's tools, turn tool/function results into assistant
+// text, and inline assistant tool_calls names instead of the structured field.
+function flattenToolHistory(messages) {
+  return messages
+    .filter((msg) => msg)
+    .map((msg) => {
+      if (msg.role === "tool" || msg.role === "function") {
+        return { role: "assistant", content: `${TOOL_RESULT_PREFIX}${extractTextContent(msg.content) || String(msg.content ?? "")}]` };
+      }
+      if (msg.role === "assistant" && Array.isArray(msg.tool_calls)) {
+        const { tool_calls, ...rest } = msg;
+        const names = tool_calls.map((c) => c?.function?.name || c?.name || "tool").join(", ");
+        const base = extractTextContent(rest.content) || (typeof rest.content === "string" ? rest.content : "");
+        return { ...rest, content: `${base}${base ? "\n" : ""}${TOOL_CALL_PREFIX}${names}]` };
+      }
+      if (Array.isArray(msg.content)) {
+        const hasToolUse = msg.content.some((c) => c.type === "tool_use");
+        const hasToolResult = msg.content.some((c) => c.type === "tool_result");
+        if (hasToolUse || hasToolResult) {
+          const textParts = [];
+          const toolNames = [];
+          const toolResults = [];
+          for (const block of msg.content) {
+            if (block.type === "text" && block.text) textParts.push(block.text);
+            if (block.type === "tool_use") toolNames.push(block.name || "tool");
+            if (block.type === "tool_result") toolResults.push(extractTextContent(block.content) || String(block.content ?? ""));
+          }
+          const { ...rest } = msg;
+          let newContent = textParts.join("\n");
+          if (toolNames.length > 0) {
+            newContent = `${newContent}${newContent ? "\n" : ""}${TOOL_CALL_PREFIX}${toolNames.join(", ")}]`;
+          }
+          if (toolResults.length > 0) {
+            newContent = `${newContent}${newContent ? "\n" : ""}${TOOL_RESULT_PREFIX}${toolResults.join("\n")}]`;
+          }
+          return { ...rest, content: newContent };
+        }
+      }
+      return msg;
+    });
+}
+
+// Reorder combo models by capability fit. Stable; never drops a model (fallback intact).
+// Tier 0: satisfies all hard + all soft. Tier 1: all hard only. Tier 2: rest.
+export function reorderByCapabilities(models, required) {
+  if (!required || required.size === 0 || !Array.isArray(models) || models.length <= 1) return models;
+  const hard = [...required].filter((c) => HARD_CAPS.has(c));
+  const soft = [...required].filter((c) => !HARD_CAPS.has(c));
+
+  const tierOf = (m) => {
+    const slash = typeof m === "string" ? m.indexOf("/") : -1;
+    const provider = slash > 0 ? m.slice(0, slash) : "";
+    const model = slash > 0 ? m.slice(slash + 1) : m;
+    const caps = getCapabilitiesForModel(provider, model);
+    if (!hard.every((c) => caps[c] === true)) return 2;
+    return soft.every((c) => caps[c] === true) ? 0 : 1;
+  };
+
+  // Stable sort by tier (Array.prototype.sort is stable in modern engines).
+  return models
+    .map((m, i) => ({ m, i, t: tierOf(m) }))
+    .sort((a, b) => a.t - b.t || a.i - b.i)
+    .map((x) => x.m);
+}
+
+/**
+ * Track rotation state per combo (for round-robin strategy)
+ * @type {Map<string, { index: number, consecutiveUseCount: number }>}
+ */
+const comboRotationState = new Map();
+
+// Trailing run of items after the last assistant/model turn = the current user
+// turn. It may span several messages (e.g. text + image split across blocks),
+// so we return all of them. History media (older turns) must not pin the combo
+// to a vision model — those get stripped + placeholdered downstream instead.
+function trailingUserItems(arr) {
+  if (!Array.isArray(arr) || arr.length === 0) return [];
+  const isAssistant = (r) => r === "assistant" || r === "model";
+  let i = arr.length - 1;
+  while (i >= 0 && !isAssistant(arr[i]?.role)) i--;
+  return arr.slice(i + 1);
+}
+
+// Detect which capabilities a request needs. Modalities (vision/pdf) are scanned
+// only on the current user turn; "search" is request-wide (lives in tools).
+// Returns a Set of: "vision" | "pdf" | "search".
+export function detectRequiredCapabilities(body) {
+  const required = new Set();
+  if (!body || typeof body !== "object") return required;
+
+  const scanBlock = (b) => {
+    if (!b || typeof b !== "object") return;
+    const t = b.type;
+    if (t === "image_url" || t === "image" || t === "input_image") required.add("vision");
+    if (t === "file" || t === "document" || t === "input_file") required.add("pdf");
+    // gemini parts: inlineData/fileData carry a mime
+    const mime = b.inlineData?.mimeType || b.fileData?.mimeType;
+    if (typeof mime === "string" && mime.startsWith("image/")) required.add("vision");
+    if (mime === "application/pdf") required.add("pdf");
+  };
+
+  const scanContent = (content) => {
+    if (Array.isArray(content)) for (const b of content) scanBlock(b);
+  };
+
+  // Modalities: current user turn only (trailing user run across each known shape).
+  for (const m of trailingUserItems(body.messages)) scanContent(m.content);      // openai / claude
+  for (const it of trailingUserItems(body.input)) scanContent(it.content);       // responses
+  const contents = body.contents || body.request?.contents;                      // gemini / antigravity
+  for (const c of trailingUserItems(contents)) scanContent(c.parts);
+
+  // search: temporarily disabled in auto-switch (feature not wired yet).
+
+  return required;
+}
+
+function normalizeStickyLimit(stickyLimit) {
+  const parsed = Number.parseInt(stickyLimit, 10);
+  return Number.isFinite(parsed) && parsed > 0 ? parsed : 1;
+}
+
+function rotateModelsFromIndex(models, currentIndex) {
+  const rotatedModels = [...models];
+  for (let i = 0; i < currentIndex; i++) {
+    const moved = rotatedModels.shift();
+    rotatedModels.push(moved);
+  }
+  return rotatedModels;
+}
+
+/**
+ * Get rotated model list based on strategy
+ * @param {string[]} models - Array of model strings
+ * @param {string} comboName - Name of the combo
+ * @param {string} strategy - "fallback" or "round-robin"
+ * @param {number|string} [stickyLimit=1] - Requests per combo model before switching
+ * @returns {string[]} Rotated models array
+ */
+export function getRotatedModels(models, comboName, strategy, stickyLimit = 1) {
+  if (!models || models.length <= 1 || strategy !== "round-robin") {
+    return models;
+  }
+
+  const rotationKey = comboName || "__default__";
+  const normalizedStickyLimit = normalizeStickyLimit(stickyLimit);
+  const existingState = comboRotationState.get(rotationKey);
+  const state = typeof existingState === "number"
+    ? { index: existingState, consecutiveUseCount: 0 }
+    : (existingState || { index: 0, consecutiveUseCount: 0 });
+
+  const currentIndex = state.index % models.length;
+  const rotatedModels = rotateModelsFromIndex(models, currentIndex);
+  const nextUseCount = state.consecutiveUseCount + 1;
+
+  if (nextUseCount >= normalizedStickyLimit) {
+    comboRotationState.set(rotationKey, {
+      index: (currentIndex + 1) % models.length,
+      consecutiveUseCount: 0,
+    });
+  } else {
+    comboRotationState.set(rotationKey, {
+      index: currentIndex,
+      consecutiveUseCount: nextUseCount,
+    });
+  }
+
+  return rotatedModels;
+}
+
+/**
+ * Reset in-memory rotation state when combo/settings change
+ * @param {string} [comboName] - Combo name to reset; omit to clear all
+ */
+export function resetComboRotation(comboName) {
+  if (comboName) comboRotationState.delete(comboName);
+  else comboRotationState.clear();
+}
+
+/**
+ * Get combo models from combos data
+ * @param {string} modelStr - Model string to check
+ * @param {Array|Object} combosData - Array of combos or object with combos
+ * @returns {string[]|null} Array of models or null if not a combo
+ */
+export function getComboModelsFromData(modelStr, combosData) {
+  // Don't check if it's in provider/model format
+  if (modelStr.includes("/")) return null;
+  
+  // Handle both array and object formats
+  const combos = Array.isArray(combosData) ? combosData : (combosData?.combos || []);
+  
+  const combo = combos.find(c => c.name === modelStr);
+  if (combo && combo.models && combo.models.length > 0) {
+    return combo.models;
+  }
+  return null;
+}
+
+/**
+ * Handle combo chat with fallback
+ * @param {Object} options
+ * @param {Object} options.body - Request body
+ * @param {string[]} options.models - Array of model strings to try
+ * @param {Function} options.handleSingleModel - Function to handle single model: (body, modelStr) => Promise<Response>
+ * @param {Object} options.log - Logger object
+ * @param {string} [options.comboName] - Name of the combo (for round-robin tracking)
+ * @param {string} [options.comboStrategy] - Strategy: "fallback" or "round-robin"
+ * @param {number|string} [options.comboStickyLimit=1] - Requests per combo model before switching
+ * @returns {Promise<Response>}
+ */
+export async function handleComboChat({ body, models, handleSingleModel, log, comboName, comboStrategy, comboStickyLimit = 1, autoSwitch = true }) {
+  // Apply rotation strategy if enabled
+  let rotatedModels = getRotatedModels(models, comboName, comboStrategy, comboStickyLimit);
+
+  // Auto-switch: float models that satisfy the request's required capabilities to the front.
+  if (autoSwitch) {
+    const required = detectRequiredCapabilities(body);
+    if (required.size > 0) {
+      const reordered = reorderByCapabilities(rotatedModels, required);
+      if (reordered[0] !== rotatedModels[0]) {
+        log.info("COMBO", `auto-switch for [${[...required].join(",")}] → ${reordered[0]}`);
+      }
+      rotatedModels = reordered;
+    }
+  }
+  
+  let lastError = null;
+  let earliestRetryAfter = null;
+  let lastStatus = null;
+
+  for (let i = 0; i < rotatedModels.length; i++) {
+    const modelStr = rotatedModels[i];
+    log.info("COMBO", `Trying model ${i + 1}/${rotatedModels.length}: ${modelStr}`);
+
+    try {
+      const result = await handleSingleModel(body, modelStr);
+      
+      // Success (2xx) - return response
+      if (result.ok) {
+        log.info("COMBO", `Model ${modelStr} succeeded`);
+        return result;
+      }
+
+      // Extract error info from response
+      let errorText = result.statusText || "";
+      let retryAfter = null;
+      try {
+        const errorBody = await result.clone().json();
+        errorText = errorBody?.error?.message || errorBody?.error || errorBody?.message || errorText;
+        retryAfter = errorBody?.retryAfter || null;
+      } catch {
+        // Ignore JSON parse errors
+      }
+
+      // Track earliest retryAfter across all combo models
+      if (retryAfter && (!earliestRetryAfter || new Date(retryAfter) < new Date(earliestRetryAfter))) {
+        earliestRetryAfter = retryAfter;
+      }
+
+      // Normalize error text to string (Worker-safe)
+      if (typeof errorText !== "string") {
+        try { errorText = JSON.stringify(errorText); } catch { errorText = String(errorText); }
+      }
+
+      // Check if should fallback to next model
+      const { shouldFallback, cooldownMs } = checkFallbackError(result.status, errorText);
+
+      if (!shouldFallback) {
+        log.warn("COMBO", `Model ${modelStr} failed (no fallback)`, { status: result.status });
+        return result;
+      }
+
+      // For transient errors (503/502/504), wait for cooldown before falling through
+      // so a briefly-overloaded provider gets a chance to recover rather than being
+      // skipped immediately (fixes: combo falls through on transient 503)
+      if (cooldownMs && cooldownMs > 0 && cooldownMs <= 5000 &&
+          (result.status === 503 || result.status === 502 || result.status === 504)) {
+        log.info("COMBO", `Model ${modelStr} transient ${result.status}, waiting ${cooldownMs}ms before next`);
+        await new Promise(r => setTimeout(r, cooldownMs));
+      }
+
+      // Fallback to next model
+      lastError = errorText || String(result.status);
+      if (!lastStatus) lastStatus = result.status;
+      log.warn("COMBO", `Model ${modelStr} failed, trying next`, { status: result.status });
+    } catch (error) {
+      // Catch unexpected exceptions to ensure fallback continues
+      lastError = error.message || String(error);
+      if (!lastStatus) lastStatus = 500;
+      log.warn("COMBO", `Model ${modelStr} threw error, trying next`, { error: lastError });
+    }
+  }
+
+  // All models failed
+  // Use 503 (Service Unavailable) rather than 406 (Not Acceptable) — 406 implies
+  // the request itself is invalid, but here the providers are simply unavailable
+  // or have no active credentials. 503 is more accurate and retryable by clients.
+  const allDisabled = lastError && lastError.toLowerCase().includes("no credentials");
+  const status = allDisabled ? 503 : (lastStatus || 503);
+  const msg = lastError || "All combo models unavailable";
+
+  if (earliestRetryAfter) {
+    const retryHuman = formatRetryAfter(earliestRetryAfter);
+    log.warn("COMBO", `All models failed | ${msg} (${retryHuman})`);
+    return unavailableResponse(status, msg, earliestRetryAfter, retryHuman);
+  }
+
+  log.warn("COMBO", `All models failed | ${msg}`);
+  return new Response(
+    JSON.stringify({ error: { message: msg } }),
+    { status, headers: { "Content-Type": "application/json" } }
+  );
+}
+
+/**
+ * Extract assistant text from a non-stream completion across formats
+ * (OpenAI chat, Claude messages, Gemini, OpenAI Responses). Returns "" if none.
+ * Panel responses are already translated to the client format by chatCore, so the
+ * leaf content→string step reuses the translator's own extractTextContent.
+ */
+function extractPanelText(json) {
+  if (!json || typeof json !== "object") return "";
+
+  // OpenAI chat completion
+  const choice = json.choices?.[0];
+  if (choice) {
+    const msg = choice.message ?? choice.delta ?? {};
+    const t = extractTextContent(msg.content);
+    if (t.trim()) return t;
+    if (typeof choice.text === "string" && choice.text.trim()) return choice.text;
+  }
+
+  // Claude messages (text blocks share OpenAI's {type:"text"} shape)
+  const claudeText = extractTextContent(json.content);
+  if (claudeText.trim()) return claudeText;
+
+  // Gemini (parts carry .text without a type discriminator)
+  const parts = json.candidates?.[0]?.content?.parts;
+  if (Array.isArray(parts)) {
+    const t = parts.map((p) => p?.text || "").join("");
+    if (t.trim()) return t;
+  }
+
+  // OpenAI Responses API
+  if (Array.isArray(json.output)) {
+    const t = json.output
+      .flatMap((o) => (Array.isArray(o.content) ? o.content.map((c) => c?.text || "") : []))
+      .join("");
+    if (t.trim()) return t;
+  }
+
+  return "";
+}
+
+/**
+ * Append a synthesized user turn to whichever message array the request format uses.
+ * Preserves the original conversation + system prompt so the judge has full context.
+ */
+function appendUserTurn(body, text) {
+  const next = { ...body };
+  if (Array.isArray(body.messages)) {
+    next.messages = [...body.messages, { role: "user", content: text }];
+  } else if (Array.isArray(body.input)) {
+    next.input = [...body.input, { role: "user", content: text }];
+  } else if (Array.isArray(body.contents)) {
+    next.contents = [...body.contents, { role: "user", parts: [{ text }] }];
+  } else {
+    next.messages = [{ role: "user", content: text }];
+  }
+  return next;
+}
+
+/**
+ * Build the judge directive. Per OpenRouter's Fusion design, the judge does NOT
+ * merge — it analyzes (consensus / contradictions / partial coverage / unique
+ * insights / blind spots) then writes one answer grounded in that analysis.
+ * ~3/4 of fusion's quality lift comes from this synthesis step.
+ *
+ * Sources are anonymized ("Source N") so the judge weighs substance, not the
+ * reputation of a model brand.
+ */
+function buildJudgePrompt(answers) {
+  const panel = answers
+    .map((a, i) => `[Source ${i + 1}]\n${a.text}`)
+    .join("\n\n");
+
+  return [
+    `You are the JUDGE in a model-fusion panel. ${answers.length} expert models independently answered the user's most recent request. Their responses are below, anonymized by source.`,
+    "",
+    "Do NOT mention that multiple models were used, and do NOT refer to the sources. Produce ONE authoritative final answer addressed directly to the user.",
+    "",
+    "First, internally analyze the panel along these dimensions: consensus (points most sources agree on — treat as higher-confidence), contradictions (where they disagree — resolve with your own judgment), partial coverage, unique insights only one source surfaced, and blind spots every source missed. Then write the best possible final answer grounded in that analysis — more complete and correct than any single response, with no filler.",
+    "",
+    "=== PANEL RESPONSES ===",
+    panel,
+    "=== END PANEL RESPONSES ===",
+    "",
+    "Now write the final answer to the user's original request.",
+  ].join("\n");
+}
+
+// Fusion tuning. Overridable per-combo via settings.comboStrategies[name].
+const FUSION_DEFAULTS = {
+  minPanel: 2,             // answers needed before stragglers get a grace window
+  stragglerGraceMs: 8000,  // wait this long for laggards once quorum is reached
+  panelHardTimeoutMs: 90000, // absolute cap so one hung model can't stall forever
+};
+
+// Resolve a Response (or {__error}) within ms; the loser keeps running but is ignored.
+function withTimeout(promise, ms) {
+  return new Promise((resolve) => {
+    const t = setTimeout(() => resolve({ __timeout: true }), ms);
+    Promise.resolve(promise)
+      .then((v) => { clearTimeout(t); resolve(v); })
+      .catch((e) => { clearTimeout(t); resolve({ __error: e }); });
+  });
+}
+
+/**
+ * Collect panel responses with quorum-grace: as soon as `minPanel` calls succeed,
+ * start a short grace timer for the rest, then proceed with whatever arrived. This
+ * caps the straggler penalty (the slowest model otherwise dominates wall time) while
+ * still preferring a full panel when everyone is fast. Bounded by a hard timeout.
+ * Returns a sparse array aligned to `calls` (undefined = not yet / dropped).
+ */
+function collectPanel(calls, { minPanel, stragglerGraceMs, panelHardTimeoutMs }) {
+  return new Promise((resolve) => {
+    const out = new Array(calls.length);
+    let settled = 0;
+    let ok = 0;
+    let finished = false;
+    let graceTimer = null;
+    const finish = () => {
+      if (finished) return;
+      finished = true;
+      clearTimeout(hardTimer);
+      if (graceTimer) clearTimeout(graceTimer);
+      resolve(out);
+    };
+    const hardTimer = setTimeout(finish, panelHardTimeoutMs);
+    calls.forEach((p, i) => {
+      Promise.resolve(p)
+        .then((v) => { out[i] = v; })
+        .catch((e) => { out[i] = { __error: e }; })
+        .finally(() => {
+          settled++;
+          if (out[i] && out[i].ok) ok++;
+          if (settled === calls.length) return finish();
+          if (ok >= minPanel && !graceTimer) graceTimer = setTimeout(finish, stragglerGraceMs);
+        });
+    });
+  });
+}
+
+/**
+ * Handle a fusion combo: fan the prompt out to every panel model in parallel,
+ * then a judge model synthesizes one final answer from all panel responses.
+ *
+ * Panel calls are forced non-streaming with tools stripped (the judge needs
+ * complete prose to synthesize). The judge call keeps the client's original
+ * stream flag + tools, so streaming and downstream tool use still work.
+ *
+ * Speed: quorum-grace collection caps the straggler penalty. Quality: the judge
+ * runs the consensus/contradiction/blind-spot analysis before writing.
+ *
+ * Degrades gracefully: 0 panel answers -> 503, exactly 1 -> return it directly.
+ *
+ * @param {Object} options
+ * @param {Object} options.body - Request body (client format)
+ * @param {string[]} options.models - Panel model strings
+ * @param {Function} options.handleSingleModel - (body, modelStr) => Promise<Response>
+ * @param {Object} options.log - Logger
+ * @param {string} [options.comboName] - Combo name (logging)
+ * @param {string} [options.judgeModel] - Judge model; falls back to panel[0]
+ * @param {Object} [options.tuning] - Override FUSION_DEFAULTS (minPanel, grace, timeout)
+ * @returns {Promise<Response>}
+ */
+export async function handleFusionChat({ body, models, handleSingleModel, log, comboName, judgeModel, tuning }) {
+  const panel = Array.isArray(models) ? models.filter(Boolean) : [];
+  if (panel.length === 0) {
+    return new Response(
+      JSON.stringify({ error: { message: "Fusion combo has no models" } }),
+      { status: 400, headers: { "Content-Type": "application/json" } }
+    );
+  }
+
+  // A single-model fusion has nothing to fuse — just answer directly.
+  if (panel.length === 1) {
+    return handleSingleModel(body, panel[0]);
+  }
+
+  const cfg = { ...FUSION_DEFAULTS, ...(tuning || {}) };
+  const minPanel = Math.min(Math.max(2, cfg.minPanel), panel.length);
+  const judge = judgeModel && judgeModel.trim() ? judgeModel.trim() : panel[0];
+  log.info("FUSION", `Combo "${comboName}" | panel=${panel.length} [${panel.join(", ")}] | judge=${judge} | quorum=${minPanel}`);
+
+  // 1. Fan out to the panel in parallel: non-streaming, tools stripped (we want prose).
+  const { tools, tool_choice, ...rest } = body;
+  const panelBody = { ...rest, stream: false };
+
+  // Flatten tool turns to prose so panel models keep context without emitting tool_calls.
+  if (Array.isArray(panelBody.messages)) {
+    panelBody.messages = flattenToolHistory(panelBody.messages);
+  } else if (Array.isArray(panelBody.input)) {
+    panelBody.input = flattenToolHistory(panelBody.input);
+  }
+
+  const t0 = Date.now();
+  const calls = panel.map((m) => withTimeout(handleSingleModel(panelBody, m, true), cfg.panelHardTimeoutMs));
+  const settled = await collectPanel(calls, { ...cfg, minPanel });
+  log.info("FUSION", `fan-out collected in ${Date.now() - t0}ms`);
+
+  // 2. Collect successful answers.
+  const answers = [];
+  for (let i = 0; i < settled.length; i++) {
+    const res = settled[i];
+    const model = panel[i];
+    if (!res) { log.warn("FUSION", `Panel ${model} dropped (straggler/timeout)`); continue; }
+    if (res.__timeout) { log.warn("FUSION", `Panel ${model} timed out`); continue; }
+    if (res.__error) { log.warn("FUSION", `Panel ${model} threw`, { error: res.__error?.message || String(res.__error) }); continue; }
+    if (!res.ok) { log.warn("FUSION", `Panel ${model} failed`, { status: res.status }); continue; }
+    try {
+      const json = await res.clone().json();
+      const text = extractPanelText(json);
+      if (text) {
+        answers.push({ model, text });
+        log.info("FUSION", `Panel ${model} ok (${text.length} chars)`);
+      } else {
+        log.warn("FUSION", `Panel ${model} returned empty content`);
+      }
+    } catch (e) {
+      log.warn("FUSION", `Panel ${model} unparseable`, { error: e.message || String(e) });
+    }
+  }
+
+  // 3. Degrade gracefully when the panel is too thin to fuse.
+  if (answers.length === 0) {
+    log.warn("FUSION", "All panel models failed");
+    return new Response(
+      JSON.stringify({ error: { message: "All fusion panel models failed" } }),
+      { status: 503, headers: { "Content-Type": "application/json" } }
+    );
+  }
+  if (answers.length === 1) {
+    log.info("FUSION", `Only ${answers[0].model} succeeded — answering directly (no fusion)`);
+    return handleSingleModel(body, answers[0].model);
+  }
+
+  // 4. Judge analyzes + writes one final answer (streams to client if requested).
+  const judgeBody = appendUserTurn(body, buildJudgePrompt(answers));
+  log.info("FUSION", `Judging ${answers.length} answers with ${judge}`);
+  return handleSingleModel(judgeBody, judge);
+}

package/open-sse/services/comboAgentMiddleware.js

@@ -0,0 +1,91 @@
+const CACHE_TAG_PATTERN = /<omniModel>([^<]+)<\/omniModel>/;
+const CACHE_TAG_PATTERN_GLOBAL = /(?:\\n|\n|\r){0,16}<omniModel>([^<]+)<\/omniModel>(?:\\n|\n|\r){0,16}/g;
+function injectModelTag(messages, providerModel) {
+  const cleaned = messages.map((msg2) => {
+    if (msg2.role === "assistant" && typeof msg2.content === "string") {
+      return { ...msg2, content: msg2.content.replace(CACHE_TAG_PATTERN_GLOBAL, "").trimEnd() };
+    }
+    return msg2;
+  });
+  const lastAssistantIdx = cleaned.map((m) => m.role).lastIndexOf("assistant");
+  if (lastAssistantIdx === -1) {
+    return [...cleaned, { role: "assistant", content: `<omniModel>${providerModel}</omniModel>` }];
+  }
+  const msg = cleaned[lastAssistantIdx];
+  if (typeof msg.content !== "string") {
+    return [...cleaned, { role: "assistant", content: `<omniModel>${providerModel}</omniModel>` }];
+  }
+  const tagged = [...cleaned];
+  tagged[lastAssistantIdx] = {
+    ...msg,
+    content: `${msg.content}<omniModel>${providerModel}</omniModel>`
+  };
+  return tagged;
+}
+function extractPinnedModel(messages) {
+  for (let i = messages.length - 1; i >= 0; i--) {
+    const msg = messages[i];
+    if (msg.role === "assistant" && typeof msg.content === "string") {
+      const match = CACHE_TAG_PATTERN.exec(msg.content);
+      if (match) return match[1];
+    }
+  }
+  return null;
+}
+function applySystemMessageOverride(messages, systemMessage) {
+  const filtered = messages.filter((m) => m.role !== "system");
+  return [{ role: "system", content: systemMessage }, ...filtered];
+}
+function applyToolFilter(tools, pattern) {
+  if (!tools || !pattern) return tools;
+  let regex;
+  try {
+    regex = new RegExp(pattern);
+  } catch {
+    console.warn(`[ComboAgent] Invalid tool_filter_regex: "${pattern}"`);
+    return tools;
+  }
+  return tools.filter((tool) => {
+    const t = tool;
+    const name = t.function?.name ?? t.name ?? "";
+    return regex.test(String(name));
+  });
+}
+function stripModelTags(messages) {
+  return messages.map((msg) => {
+    if (typeof msg.content === "string" && CACHE_TAG_PATTERN.test(msg.content)) {
+      return { ...msg, content: msg.content.replace(CACHE_TAG_PATTERN_GLOBAL, "").trimEnd() };
+    }
+    return msg;
+  });
+}
+function applyComboAgentMiddleware(body, comboConfig, providerModel) {
+  if (!comboConfig) return { body, pinnedModel: null };
+  let messages = Array.isArray(body.messages) ? [...body.messages] : [];
+  let pinnedModel = null;
+  pinnedModel = null;
+  if (comboConfig.system_message && comboConfig.system_message.trim()) {
+    messages = applySystemMessageOverride(messages, comboConfig.system_message);
+  }
+  const filteredTools = applyToolFilter(
+    body.tools,
+    comboConfig.tool_filter_regex
+  );
+  messages = stripModelTags(messages);
+  return {
+    body: {
+      ...body,
+      messages,
+      ...filteredTools !== body.tools && { tools: filteredTools }
+    },
+    pinnedModel
+  };
+}
+export {
+  applyComboAgentMiddleware,
+  applySystemMessageOverride,
+  applyToolFilter,
+  extractPinnedModel,
+  injectModelTag,
+  stripModelTags
+};