zevairouter 1.0.154 → 1.0.156

package/open-sse/config/kiroConstants.js

@@ -1,360 +0,0 @@
-/**
- * Kiro-specific constants and helpers.
- *
- * Mirrors the behaviour of `internal/translator/kiro/common/constants.go` and
- * `internal/translator/kiro/claude/kiro_claude_request.go` from the
- * CLIProxyAPIPlus reference implementation, scoped down to what zevairouter needs:
- *
- *   - `-agentic` model suffix detection + chunked-write system prompt
- *   - reasoning / thinking trigger detection (Anthropic-Beta header,
- *     Claude `thinking`, OpenAI `reasoning_effort`, AMP/Cursor magic tag)
- *   - the `<thinking_mode>enabled</thinking_mode>` system-prompt injection
- *     that turns Kiro reasoning on
- *
- * Kiro upstream does not advertise `-agentic` model IDs; they are a zevairouter
- * fiction. The suffix is stripped before the request leaves this process.
- */
-
-export const KIRO_AGENTIC_SUFFIX = "-agentic";
-export const KIRO_THINKING_SUFFIX = "-thinking";
-
-export const KIRO_THINKING_BUDGET_DEFAULT = 16000;
-
-// Client fingerprints presented to Kiro/CodeWhisperer APIs (match current IDE/CLI).
-// IDE 1.0.138 (Jul 2026), CLI 2.12.2 (2.12 line Jul 2026).
-export const KIRO_IDE_VERSION = "1.0.138";
-export const KIRO_CLI_VERSION = "2.12.2";
-export const KIRO_RUNTIME_SDK_VERSION = "1.0.0";
-export const KIRO_NODE_VERSION = "22.21.1";
-export const KIRO_AGENT_OS = "windows";
-export const KIRO_AGENT_OS_VERSION = "10.0.26200";
-
-/** User-Agent for kiro-cli-style token refresh / social auth. */
-export function kiroCliUserAgent() {
-  return `kiro-cli/${KIRO_CLI_VERSION}`;
-}
-
-/** Headers for generateAssistantResponse-style streaming requests. */
-export function kiroIdeRequestHeaders() {
-  return {
-    "User-Agent": `AWS-SDK-JS/3.0.0 kiro-ide/${KIRO_IDE_VERSION}`,
-    "X-Amz-User-Agent": `aws-sdk-js/3.0.0 kiro-ide/${KIRO_IDE_VERSION}`,
-  };
-}
-
-/**
- * Per-account fingerprint headers for ListAvailableModels (must match Kiro IDE UA shape).
- * @param {string} machineId hex machine id
- */
-export function buildKiroIdeRuntimeUserAgents(machineId) {
-  const userAgent =
-    `aws-sdk-js/${KIRO_RUNTIME_SDK_VERSION} ua/2.1 ` +
-    `os/${KIRO_AGENT_OS}#${KIRO_AGENT_OS_VERSION} ` +
-    `lang/js md/nodejs#${KIRO_NODE_VERSION} ` +
-    `api/codewhispererruntime#${KIRO_RUNTIME_SDK_VERSION} m/N,E ` +
-    `KiroIDE-${KIRO_IDE_VERSION}-${machineId}`;
-  const amzUserAgent =
-    `aws-sdk-js/${KIRO_RUNTIME_SDK_VERSION} KiroIDE-${KIRO_IDE_VERSION}-${machineId}`;
-  return { userAgent, amzUserAgent };
-}
-
-const TRUE_VALUES = new Set(["1", "true", "yes", "on"]);
-const FALSE_VALUES = new Set(["0", "false", "no", "off"]);
-
-export const KIRO_AGENTIC_SYSTEM_PROMPT = `
-# CRITICAL: CHUNKED WRITE PROTOCOL (MANDATORY)
-
-You MUST follow these rules for ALL file operations. Violation causes server timeouts and task failure.
-
-## ABSOLUTE LIMITS
-- **MAXIMUM 350 LINES** per single write/edit operation - NO EXCEPTIONS
-- **RECOMMENDED 300 LINES** or less for optimal performance
-- **NEVER** write entire files in one operation if >300 lines
-
-## MANDATORY CHUNKED WRITE STRATEGY
-
-### For NEW FILES (>300 lines total):
-1. FIRST: Write initial chunk (first 250-300 lines) using write_to_file/fsWrite
-2. THEN: Append remaining content in 250-300 line chunks using file append operations
-3. REPEAT: Continue appending until complete
-
-### For EDITING EXISTING FILES:
-1. Use surgical edits (apply_diff/targeted edits) - change ONLY what's needed
-2. NEVER rewrite entire files - use incremental modifications
-3. Split large refactors into multiple small, focused edits
-
-### For LARGE CODE GENERATION:
-1. Generate in logical sections (imports, types, functions separately)
-2. Write each section as a separate operation
-3. Use append operations for subsequent sections
-
-## EXAMPLES OF CORRECT BEHAVIOR
-
-CORRECT: Writing a 600-line file
-- Operation 1: Write lines 1-300 (initial file creation)
-- Operation 2: Append lines 301-600
-
-CORRECT: Editing multiple functions
-- Operation 1: Edit function A
-- Operation 2: Edit function B
-- Operation 3: Edit function C
-
-WRONG: Writing 500 lines in single operation -> TIMEOUT
-WRONG: Rewriting entire file to change 5 lines -> TIMEOUT
-WRONG: Generating massive code blocks without chunking -> TIMEOUT
-
-## WHY THIS MATTERS
-- Server has 2-3 minute timeout for operations
-- Large writes exceed timeout and FAIL completely
-- Chunked writes are FASTER and more RELIABLE
-- Failed writes waste time and require retry
-
-REMEMBER: When in doubt, write LESS per operation. Multiple small operations > one large operation.
-`.trim();
-
-/**
- * Detect whether an inbound request is asking for reasoning / thinking output.
- *
- * Sources of intent (any one is enough):
- *   - HTTP header `Anthropic-Beta: ...interleaved-thinking...`
- *   - JSON `thinking.type === "enabled"` (Claude Messages API)
- *   - JSON `reasoning_effort` in {low, medium, high, auto} (OpenAI o1/o3)
- *   - JSON `reasoning.effort` in {low, medium, high, auto} (OpenAI Responses)
- *   - System prompt contains `<thinking_mode>enabled</thinking_mode>` or
- *     `<thinking_mode>interleaved</thinking_mode>` (AMP / Cursor)
- *   - Model name contains `thinking` or `-reason`
- *
- * @param {object} body OpenAI-shaped request body (post-translation)
- * @param {object} [headers] Original inbound HTTP headers (case-insensitive)
- * @param {string} [model] Model id the caller asked for (post-strip ok)
- * @returns {boolean}
- */
-export function isThinkingEnabled(body, headers, model) {
-  if (isThinkingExplicitlyDisabled(body)) {
-    return false;
-  }
-
-  if (headers) {
-    const beta = pickHeader(headers, "anthropic-beta");
-    if (typeof beta === "string" && beta.toLowerCase().includes("interleaved-thinking")) {
-      return true;
-    }
-  }
-
-  if (body && typeof body === "object") {
-    const thinking = body.thinking;
-    if (thinking && typeof thinking === "object" && thinking.type === "enabled") {
-      const budget = Number(thinking.budget_tokens);
-      if (!Number.isFinite(budget) || budget > 0) {
-        return true;
-      }
-    }
-
-    const effort = body.reasoning_effort
-      ?? (body.reasoning && typeof body.reasoning === "object" ? body.reasoning.effort : null);
-    if (typeof effort === "string") {
-      const v = effort.toLowerCase();
-      if (v && v !== "none" && (v === "low" || v === "medium" || v === "high" || v === "auto")) {
-        return true;
-      }
-    }
-
-    if (containsThinkingModeTag(body)) {
-      return true;
-    }
-  }
-
-  if (typeof model === "string" && model) {
-    const m = model.toLowerCase();
-    if (m.includes("thinking") || m.includes("-reason")) {
-      return true;
-    }
-  }
-
-  return false;
-}
-
-function isThinkingExplicitlyDisabled(body) {
-  if (!body || typeof body !== "object") return false;
-  const thinking = body.thinking;
-  if (thinking && typeof thinking === "object" && thinking.type === "disabled") {
-    return true;
-  }
-  const effort = body.reasoning_effort
-    ?? (body.reasoning && typeof body.reasoning === "object" ? body.reasoning.effort : null);
-  return typeof effort === "string" && effort.toLowerCase() === "none";
-}
-
-/**
- * Decide whether Kiro reasoning chunks should be forwarded to the client.
- *
- * Kiro's reasoning stream is useful for debugging, but most downstream clients
- * display it as user-visible "thinking" progress. Keep the default quiet and
- * let power users opt in per request or by environment variable.
- */
-export function shouldExposeKiroReasoning(body) {
-  const explicit = firstDefined(
-    body?.kiro_expose_reasoning,
-    body?.kiroExposeReasoning,
-    body?.metadata?.kiro_expose_reasoning,
-    body?.metadata?.kiroExposeReasoning,
-    body?.extra_body?.kiro_expose_reasoning,
-    body?.extra_body?.kiroExposeReasoning
-  );
-  const parsedExplicit = parseBooleanLike(explicit);
-  if (parsedExplicit !== null) return parsedExplicit;
-
-  const envSource = typeof process !== "undefined" ? process.env : {};
-  const env = firstDefined(
-    envSource.KIRO_EXPOSE_REASONING,
-    envSource.ZEVAI_EXPOSE_REASONING
-  );
-  const parsedEnv = parseBooleanLike(env);
-  return parsedEnv === true;
-}
-
-/**
- * Detect whether a model id refers to a zevairouter synthetic agentic variant.
- * Agentic variants share the same upstream model as the base; the only
- * difference is the chunked-write system prompt this module injects.
- *
- * @param {string} model
- * @returns {boolean}
- */
-export function isAgenticModel(model) {
-  return typeof model === "string" && model.endsWith(KIRO_AGENTIC_SUFFIX);
-}
-
-/**
- * Strip the `-agentic` suffix from a model id, leaving the upstream-real id.
- *
- * @param {string} model
- * @returns {string}
- */
-export function stripAgenticSuffix(model) {
-  if (!isAgenticModel(model)) return model;
-  return model.slice(0, -KIRO_AGENTIC_SUFFIX.length);
-}
-
-/**
- * Detect whether a model id is a zevairouter synthetic thinking variant
- * (e.g. `claude-sonnet-4.5-thinking`). Same upstream model as the base; the
- * only difference is `<thinking_mode>enabled</thinking_mode>` injection.
- *
- * Note: real Kiro thinking-capable variants exist (e.g. `kimi-k2-thinking` in
- * other providers), but for the `kr/` namespace there is no `-thinking`
- * model on Kiro upstream. Treat the suffix as a synthetic alias.
- *
- * @param {string} model Model id with `-agentic` already stripped
- * @returns {boolean}
- */
-export function isThinkingModel(model) {
-  return typeof model === "string" && model.endsWith(KIRO_THINKING_SUFFIX);
-}
-
-/**
- * Strip the `-thinking` suffix from a model id.
- *
- * @param {string} model
- * @returns {string}
- */
-export function stripThinkingSuffix(model) {
-  if (!isThinkingModel(model)) return model;
-  return model.slice(0, -KIRO_THINKING_SUFFIX.length);
-}
-
-/**
- * Resolve a zevairouter model id to the real upstream Kiro model id, plus flags
- * describing which behaviours the suffixes implied.
- *
- *   resolveKiroModel("claude-sonnet-4.5-thinking-agentic")
- *     => { upstream: "claude-sonnet-4.5", agentic: true, thinking: true }
- *   resolveKiroModel("claude-sonnet-4.5-thinking")
- *     => { upstream: "claude-sonnet-4.5", agentic: false, thinking: true }
- *   resolveKiroModel("claude-sonnet-4.5-agentic")
- *     => { upstream: "claude-sonnet-4.5", agentic: true, thinking: false }
- *   resolveKiroModel("claude-sonnet-4.5")
- *     => { upstream: "claude-sonnet-4.5", agentic: false, thinking: false }
- *
- * @param {string} model
- * @returns {{ upstream: string, agentic: boolean, thinking: boolean }}
- */
-export function resolveKiroModel(model) {
-  let upstream = model;
-  let agentic = false;
-  let thinking = false;
-  if (isAgenticModel(upstream)) {
-    agentic = true;
-    upstream = stripAgenticSuffix(upstream);
-  }
-  if (isThinkingModel(upstream)) {
-    thinking = true;
-    upstream = stripThinkingSuffix(upstream);
-  }
-  return { upstream, agentic, thinking };
-}
-
-/**
- * Build the magic system-prompt prefix that turns Kiro reasoning on.
- * Same shape as CLIProxyAPIPlus.
- *
- * @param {number} [budget=KIRO_THINKING_BUDGET_DEFAULT]
- */
-export function buildThinkingSystemPrefix(budget = KIRO_THINKING_BUDGET_DEFAULT) {
-  const safeBudget = Math.max(1, Math.min(32000, Number(budget) || KIRO_THINKING_BUDGET_DEFAULT));
-  return `<thinking_mode>enabled</thinking_mode>\n<max_thinking_length>${safeBudget}</max_thinking_length>`;
-}
-
-function pickHeader(headers, name) {
-  if (!headers) return undefined;
-  if (typeof headers.get === "function") {
-    return headers.get(name);
-  }
-  const lower = name.toLowerCase();
-  for (const key of Object.keys(headers)) {
-    if (key.toLowerCase() === lower) {
-      return headers[key];
-    }
-  }
-  return undefined;
-}
-
-function containsThinkingModeTag(body) {
-  const messages = Array.isArray(body?.messages) ? body.messages : [];
-  for (const msg of messages) {
-    if (!msg) continue;
-    if (msg.role !== "system" && msg.role !== "user") continue;
-    const content = msg.content;
-    if (typeof content === "string") {
-      if (containsTagInText(content)) return true;
-    } else if (Array.isArray(content)) {
-      for (const part of content) {
-        const text = part?.text;
-        if (typeof text === "string" && containsTagInText(text)) return true;
-      }
-    }
-  }
-  if (typeof body?.system === "string" && containsTagInText(body.system)) return true;
-  return false;
-}
-
-function containsTagInText(text) {
-  if (!text) return false;
-  if (!text.includes("<thinking_mode>")) return false;
-  return text.includes("<thinking_mode>enabled</thinking_mode>")
-    || text.includes("<thinking_mode>interleaved</thinking_mode>");
-}
-
-function firstDefined(...values) {
-  return values.find(value => value !== undefined && value !== null);
-}
-
-function parseBooleanLike(value) {
-  if (typeof value === "boolean") return value;
-  if (typeof value === "number") return value === 1 ? true : value === 0 ? false : null;
-  if (typeof value !== "string") return null;
-  const normalized = value.trim().toLowerCase();
-  if (TRUE_VALUES.has(normalized)) return true;
-  if (FALSE_VALUES.has(normalized)) return false;
-  return null;
-}

package/open-sse/config/models.js

@@ -1,13 +0,0 @@
-// Model metadata registry
-// Only define models that differ from DEFAULT_MODEL_INFO
-// Custom entries are merged over default
-const DEFAULT_MODEL_INFO = {
-  type: ["chat"],
-  contextWindow: 200000,
-};
-
-export const MODEL_INFO = {};
-
-export function getModelInfo(modelId) {
-  return { ...DEFAULT_MODEL_INFO, ...MODEL_INFO[modelId] };
-}

package/open-sse/config/ollamaModels.js

@@ -1,19 +0,0 @@
-export const ollamaModels = {
-  models: [
-    {
-      name: "llama3.2",
-      modified_at: "2025-12-26T00:00:00Z",
-      size: 2000000000,
-      digest: "abc123def456",
-      details: { format: "gguf", family: "llama", parameter_size: "3B", quantization_level: "Q4_K_M" }
-    },
-    {
-      name: "qwen2.5",
-      modified_at: "2025-12-26T00:00:00Z",
-      size: 4000000000,
-      digest: "def456abc123",
-      details: { format: "gguf", family: "qwen", parameter_size: "7B", quantization_level: "Q4_K_M" }
-    }
-  ]
-};
-