@gamaze/hicortex 0.4.3 → 0.4.5

package/dist/index.js

@@ -42,7 +42,9 @@ var __importStar = (this && this.__importStar) || (function () {
 Object.defineProperty(exports, "__esModule", { value: true });
 const node_path_1 = require("node:path");
 const db_js_1 = require("./db.js");
-const license_js_1 = require("./license.js");
+const features_js_1 = require("./features.js");
+const extensions_js_1 = require("./extensions.js");
+const state_js_1 = require("./state.js");
 const llm_js_1 = require("./llm.js");
 const node_fs_1 = require("node:fs");
 const node_os_1 = require("node:os");
@@ -87,8 +89,10 @@ exports.default = {
                 llm = new llm_js_1.LlmClient(llmConfig);
                 log(`[hicortex] LLM: ${llmConfig.provider}/${llmConfig.model} ` +
                     `(reflect: ${llmConfig.reflectModel})`);
-                // License check (non-blocking)
-                (0, license_js_1.validateLicense)(config.licenseKey, stateDir).catch((err) => log(`[hicortex] License validation failed: ${err}`));
+                // One-time migration of legacy state files (no-op if state.json exists)
+                (0, state_js_1.migrateLegacyState)(stateDir);
+                // License: init feature cache (sync after this returns)
+                await (0, features_js_1.initFeatures)(config.licenseKey, stateDir);
                 // Schedule nightly consolidation
                 const consolidateHour = config.consolidateHour ?? 2;
                 cancelConsolidation = (0, consolidate_js_1.scheduleConsolidation)(db, llm, embedder_js_1.embed, consolidateHour);
@@ -123,9 +127,13 @@ exports.default = {
                 const lessons = storage.getLessons(db, 7, ctx.project);
                 if (lessons.length === 0)
                     return {};
-                const maxFeatures = (0, license_js_1.getFeatures)(stateDir);
-                const maxLessons = maxFeatures.maxMemories === -1 ? 20 : 10;
-                const formatted = lessons.slice(0, maxLessons).map((l) => {
+                const maxLessons = (0, features_js_1.lessonsLimit)();
+                const selected = await (0, extensions_js_1.getLessonSelector)().select(lessons, {
+                    maxLessons,
+                    project: ctx.project,
+                    agentId: ctx.agentId,
+                });
+                const formatted = selected.map((l) => {
                     // Extract just the lesson text from the structured content
                     const match = l.content.match(/## Lesson: (.+)/);
                     return match ? `- ${match[1]}` : `- ${l.content.slice(0, 200)}`;
@@ -136,9 +144,9 @@ exports.default = {
                     "\n";
                 // Daily friendly reminder when at memory cap
                 const memCount = storage.countMemories(db);
-                if (maxFeatures.maxMemories > 0 && memCount >= maxFeatures.maxMemories) {
+                if ((0, features_js_1.memoryCapReached)(memCount)) {
                     context +=
-                        `\n---\nHicortex free tier: ${maxFeatures.maxMemories} memories stored. ` +
+                        `\n---\nHicortex free tier: ${(0, features_js_1.maxMemoriesAllowed)()} memories stored. ` +
                             `New memories can no longer be saved, and your agent has stopped learning and self-improving from new sessions. ` +
                             `Existing memories, lessons, and search still work. ` +
                             `Upgrade for unlimited usage: https://hicortex.gamaze.com/ ` +
@@ -169,10 +177,8 @@ exports.default = {
                 if (entries.length === 0)
                     return;
                 // Check license cap
-                const features = (0, license_js_1.getFeatures)(stateDir);
-                if (features.maxMemories > 0 &&
-                    storage.countMemories(db) >= features.maxMemories) {
-                    console.warn(`[hicortex] Free tier limit reached (${features.maxMemories} memories). ` +
+                if ((0, features_js_1.memoryCapReached)(storage.countMemories(db))) {
+                    console.warn(`[hicortex] Free tier limit reached (${(0, features_js_1.maxMemoriesAllowed)()} memories). ` +
                         `Search and lessons still work, but new memories won't be saved. ` +
                         `Upgrade for unlimited usage: https://hicortex.gamaze.com/`);
                     return;
@@ -299,12 +305,11 @@ exports.default = {
             async execute(_callId, args, context) {
                 if (!db)
                     return { error: "Hicortex not initialized" };
-                const features = (0, license_js_1.getFeatures)(stateDir);
-                if (features.maxMemories > 0 && storage.countMemories(db) >= features.maxMemories) {
+                if ((0, features_js_1.memoryCapReached)(storage.countMemories(db))) {
                     return {
                         content: [{
                                 type: "text",
-                                text: `Free tier limit reached (${features.maxMemories} memories). ` +
+                                text: `Free tier limit reached (${(0, features_js_1.maxMemoriesAllowed)()} memories). ` +
                                     `Your existing memories and lessons still work — search and recall are unaffected. ` +
                                     `New memories won't be saved until you upgrade.\n\n` +
                                     `Upgrade for unlimited usage: https://hicortex.gamaze.com/`
@@ -461,28 +466,7 @@ async function autoConfigureLlm(pluginConfig, log) {
         const msg = err instanceof Error ? err.message : String(err);
         log(`[hicortex] LLM test failed (${llmConfig.baseUrl}): ${msg}`);
     }
-    // Step 4: If test failed and this is z.ai, try the alternate endpoint
-    if (llmConfig.provider === "zai") {
-        const altUrl = llmConfig.baseUrl.includes("/coding/")
-            ? llmConfig.baseUrl.replace("/coding/", "/")
-            : llmConfig.baseUrl.replace("/api/paas/", "/api/coding/paas/");
-        log(`[hicortex] Trying alternate z.ai endpoint: ${altUrl}`);
-        llmConfig.baseUrl = altUrl;
-        const altClient = new llm_js_1.LlmClient(llmConfig);
-        try {
-            const response = await altClient.completeFast("Respond with just the word OK", 10);
-            if (response && response.length > 0) {
-                log(`[hicortex] LLM connection verified on alternate endpoint`);
-                persistProviderConfig(llmConfig, log);
-                return llmConfig;
-            }
-        }
-        catch (err) {
-            const msg = err instanceof Error ? err.message : String(err);
-            log(`[hicortex] Alternate z.ai endpoint also failed: ${msg}`);
-        }
-    }
-    // Step 5: Fall back — return the config anyway, log instructions
+    // Step 4: Fall back — return the config anyway, log instructions
     log(`[hicortex] WARNING: Could not verify LLM connection. ` +
         `Distillation and consolidation may fail. ` +
         `To fix: add models.providers.${llmConfig.provider}.baseUrl to ~/.openclaw/openclaw.json ` +
@@ -529,26 +513,6 @@ function persistProviderConfig(llmConfig, log) {
         // Non-fatal — config works in memory even if we can't persist
     }
 }
-/**
- * Track when the memory cap was first hit. Returns days since cap was reached.
- * Stores timestamp in stateDir/cap-hit.txt on first detection.
- */
-function getDaysSinceCapHit(dir) {
-    const capFile = (0, node_path_1.join)(dir, "cap-hit.txt");
-    try {
-        const ts = (0, node_fs_1.readFileSync)(capFile, "utf-8").trim();
-        const hitDate = new Date(ts);
-        return Math.floor((Date.now() - hitDate.getTime()) / (1000 * 60 * 60 * 24));
-    }
-    catch {
-        // First time hitting cap — record it
-        try {
-            (0, node_fs_1.writeFileSync)(capFile, new Date().toISOString());
-        }
-        catch { /* non-fatal */ }
-        return 0;
-    }
-}
 const HICORTEX_TOOLS = [
     "hicortex_search",
     "hicortex_context",

package/dist/init.d.ts

@@ -3,7 +3,7 @@
  *
  * Detection:
  *   1. Local HC server running (localhost:8787)
- *   2. Remote HC server (HICORTEX_SERVER_URL or bedrock:8787)
+ *   2. Remote HC server (HICORTEX_SERVER_URL — any reachable host:port)
  *   3. OC plugin installed (~/.openclaw/openclaw.json)
  *   4. CC MCP already registered (~/.claude/settings.json)
  *   5. Existing DB (~/.hicortex/ or ~/.openclaw/data/)

package/dist/init.js

@@ -4,7 +4,7 @@
  *
  * Detection:
  *   1. Local HC server running (localhost:8787)
- *   2. Remote HC server (HICORTEX_SERVER_URL or bedrock:8787)
+ *   2. Remote HC server (HICORTEX_SERVER_URL — any reachable host:port)
  *   3. OC plugin installed (~/.openclaw/openclaw.json)
  *   4. CC MCP already registered (~/.claude/settings.json)
  *   5. Existing DB (~/.hicortex/ or ~/.openclaw/data/)
@@ -185,10 +185,10 @@ When invoked with \`/learn <text>\`, store the learning in long-term memory via
 
 ## Example
 
-\`/learn z.ai API uses Bearer auth on all endpoints, not x-api-key\`
+\`/learn always check provider docs before assuming an API uses the same auth scheme as OpenAI\`
 
 Becomes a call to hicortex_ingest with:
-- content: "LEARNING: z.ai API uses Bearer auth on all three endpoints (paas, coding, anthropic). Not x-api-key. (2026-03-26)"
+- content: "LEARNING: always check provider docs before assuming an API uses the same auth scheme as OpenAI — header names and token formats vary widely (Bearer vs x-api-key vs custom)."
 - memory_type: "lesson"
 `;
     const learnPath = (0, node_path_1.join)(CC_COMMANDS_DIR, "learn.md");
@@ -402,7 +402,7 @@ async function persistLlmConfig() {
     options.push({
         label: "Other provider (requires API key)",
         save: async () => {
-            console.log("\n  Providers: Anthropic, OpenAI, Google, z.ai, OpenRouter, or any OpenAI-compatible");
+            console.log("\n  Providers: Anthropic, OpenAI, Google, OpenRouter, or any OpenAI-compatible endpoint");
             const baseUrl = await ask("  Provider base URL: ");
             if (!baseUrl) {
                 console.log("  ⚠ Cancelled.");

package/dist/license.d.ts

@@ -1,5 +1,15 @@
+/**
+ * License API client.
+ *
+ * This module is a thin wrapper over the validation HTTP endpoint at
+ * https://hicortex.gamaze.com/api/validate. Persistence and feature gating
+ * live in features.ts + state.ts; this file does NOT touch disk.
+ *
+ * Offline grace: when the API is unreachable, we fall back to the cached
+ * tier in state.json (written by features.ts on the last successful
+ * validation). If the cached tier was validated within OFFLINE_GRACE_DAYS,
+ * we treat it as still valid.
+ */
 import type { LicenseInfo } from "./types.js";
-/** Validate a license key against the Hicortex API */
+/** Validate a license key against the Hicortex API. */
 export declare function validateLicense(key: string | undefined, stateDir: string): Promise<LicenseInfo>;
-/** Get current features, using cache or free tier defaults */
-export declare function getFeatures(stateDir: string): LicenseInfo["features"];

package/dist/license.js

@@ -1,13 +1,23 @@
 "use strict";
+/**
+ * License API client.
+ *
+ * This module is a thin wrapper over the validation HTTP endpoint at
+ * https://hicortex.gamaze.com/api/validate. Persistence and feature gating
+ * live in features.ts + state.ts; this file does NOT touch disk.
+ *
+ * Offline grace: when the API is unreachable, we fall back to the cached
+ * tier in state.json (written by features.ts on the last successful
+ * validation). If the cached tier was validated within OFFLINE_GRACE_DAYS,
+ * we treat it as still valid.
+ */
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.validateLicense = validateLicense;
-exports.getFeatures = getFeatures;
-const node_fs_1 = require("node:fs");
-const node_path_1 = require("node:path");
+const state_js_1 = require("./state.js");
 const VALIDATE_URL = "https://hicortex.gamaze.com/api/validate";
 const CACHE_TTL_MS = 24 * 60 * 60 * 1000; // 24 hours
 const OFFLINE_GRACE_DAYS = 7;
-// In-memory cache
+// In-memory cache for the current process
 let cachedLicense = null;
 let cacheTimestamp = 0;
 const FREE_LICENSE = {
@@ -21,12 +31,12 @@ const FREE_LICENSE = {
         remoteIngest: true,
     },
 };
-/** Validate a license key against the Hicortex API */
+/** Validate a license key against the Hicortex API. */
 async function validateLicense(key, stateDir) {
     // No key = free tier
     if (!key)
         return FREE_LICENSE;
-    // Check in-memory cache
+    // Check in-memory cache (24h TTL)
     if (cachedLicense && Date.now() - cacheTimestamp < CACHE_TTL_MS) {
         return cachedLicense;
     }
@@ -41,57 +51,36 @@ async function validateLicense(key, stateDir) {
             throw new Error(`HTTP ${resp.status}`);
         }
         const data = (await resp.json());
-        // Cache result
         cachedLicense = data;
         cacheTimestamp = Date.now();
-        // Persist last successful validation timestamp for offline grace
-        if (data.valid) {
-            persistValidationTimestamp(stateDir);
-        }
         return data;
     }
     catch {
-        // Network failure — check offline grace period
-        return offlineFallback(key, stateDir);
-    }
-}
-/** Get current features, using cache or free tier defaults */
-function getFeatures(stateDir) {
-    if (cachedLicense)
-        return cachedLicense.features;
-    return FREE_LICENSE.features;
-}
-function persistValidationTimestamp(stateDir) {
-    try {
-        (0, node_fs_1.writeFileSync)((0, node_path_1.join)(stateDir, "license-validated.txt"), new Date().toISOString());
-    }
-    catch {
-        // Non-critical
+        // Network failure — check offline grace period via state.tier
+        return offlineFallback(stateDir);
     }
 }
-function offlineFallback(key, stateDir) {
-    const tsPath = (0, node_path_1.join)(stateDir, "license-validated.txt");
-    if (!(0, node_fs_1.existsSync)(tsPath))
+/**
+ * Offline fallback: if state.tier was validated within the grace period,
+ * return its cached features as if validation succeeded. Otherwise, free tier.
+ */
+function offlineFallback(stateDir) {
+    const persisted = (0, state_js_1.loadState)(stateDir).tier;
+    if (!persisted)
         return FREE_LICENSE;
     try {
-        const lastValidated = new Date((0, node_fs_1.readFileSync)(tsPath, "utf-8").trim());
+        const lastValidated = new Date(persisted.validatedAt);
         const daysSince = (Date.now() - lastValidated.getTime()) / (1000 * 60 * 60 * 24);
         if (daysSince <= OFFLINE_GRACE_DAYS) {
-            // Within grace period — assume last known state was valid
-            return cachedLicense ?? {
+            return {
                 valid: true,
-                tier: "pro",
-                features: {
-                    reflection: true,
-                    vectorSearch: true,
-                    maxMemories: -1,
-                    crossAgent: true,
-                },
+                tier: persisted.tier,
+                features: persisted.features,
             };
         }
     }
     catch {
-        // Corrupted file
+        // Corrupted timestamp — fall through
     }
     return FREE_LICENSE;
 }

package/dist/llm.d.ts

@@ -1,6 +1,5 @@
 /**
  * Multi-provider LLM client for consolidation and distillation.
- * Ported from hicortex/consolidate/llm.py.
  *
  * Resolution for OC adapter (resolveLlmConfig):
  *   1. Plugin config (llmBaseUrl, llmApiKey, llmModel)
@@ -10,12 +9,14 @@
  *
  * Resolution for CC adapter (resolveLlmConfigForCC):
  *   1. Explicit env vars (HICORTEX_LLM_BASE_URL + HICORTEX_LLM_API_KEY + HICORTEX_LLM_MODEL)
- *   2. ANTHROPIC_API_KEY → Haiku (cheap, CC users always have this)
- *   3. OPENAI_API_KEY → gpt-4o-mini
- *   4. GOOGLE_API_KEY → gemini-2.0-flash
- *   5. Fallback: Ollama at http://localhost:11434
+ *   2. ANTHROPIC_API_KEY → Claude Haiku (cheap, CC users always have this)
+ *   3. OPENAI_API_KEY → gpt-5.4-nano
+ *   4. GOOGLE_API_KEY → gemini-2.5-flash
+ *   5. Claude CLI fallback (uses subscription, no API key needed)
+ *   6. Fallback: Ollama at http://localhost:11434
  *
- * Supports: OpenAI, Anthropic, Google, OpenRouter, Ollama, z.ai, and 15+ more
+ * Supports any OpenAI-compatible endpoint plus first-class support for
+ * OpenAI, Anthropic, Google, Ollama, OpenRouter, and Claude CLI.
  */
 export interface LlmConfig {
     baseUrl: string;
@@ -116,7 +117,7 @@ export declare class LlmClient {
      */
     private completeOllama;
     /**
-     * Anthropic Messages API (/v1/messages). Used for Anthropic and z.ai.
+     * Anthropic Messages API (/v1/messages).
      * Auth via x-api-key header.
      */
     private completeAnthropic;

package/dist/llm.js

@@ -1,7 +1,6 @@
 "use strict";
 /**
  * Multi-provider LLM client for consolidation and distillation.
- * Ported from hicortex/consolidate/llm.py.
  *
  * Resolution for OC adapter (resolveLlmConfig):
  *   1. Plugin config (llmBaseUrl, llmApiKey, llmModel)
@@ -11,12 +10,14 @@
  *
  * Resolution for CC adapter (resolveLlmConfigForCC):
  *   1. Explicit env vars (HICORTEX_LLM_BASE_URL + HICORTEX_LLM_API_KEY + HICORTEX_LLM_MODEL)
- *   2. ANTHROPIC_API_KEY → Haiku (cheap, CC users always have this)
- *   3. OPENAI_API_KEY → gpt-4o-mini
- *   4. GOOGLE_API_KEY → gemini-2.0-flash
- *   5. Fallback: Ollama at http://localhost:11434
+ *   2. ANTHROPIC_API_KEY → Claude Haiku (cheap, CC users always have this)
+ *   3. OPENAI_API_KEY → gpt-5.4-nano
+ *   4. GOOGLE_API_KEY → gemini-2.5-flash
+ *   5. Claude CLI fallback (uses subscription, no API key needed)
+ *   6. Fallback: Ollama at http://localhost:11434
  *
- * Supports: OpenAI, Anthropic, Google, OpenRouter, Ollama, z.ai, and 15+ more
+ * Supports any OpenAI-compatible endpoint plus first-class support for
+ * OpenAI, Anthropic, Google, Ollama, OpenRouter, and Claude CLI.
  */
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.LlmClient = exports.RateLimitError = void 0;
@@ -76,7 +77,7 @@ function resolveLlmConfigForCC(overrides) {
             baseUrl: overrides.llmBaseUrl,
             apiKey: overrides.llmApiKey,
             model: overrides.llmModel ?? "claude-haiku-4-5-20251001",
-            reflectModel: overrides.reflectModel ?? overrides.llmModel ?? "claude-sonnet-4-5-20250514",
+            reflectModel: overrides.reflectModel ?? overrides.llmModel ?? "claude-sonnet-4-6",
             provider,
         };
     }
@@ -90,7 +91,7 @@ function resolveLlmConfigForCC(overrides) {
             baseUrl: hcBaseUrl,
             apiKey: hcApiKey,
             model: hcModel ?? "claude-haiku-4-5-20251001",
-            reflectModel: process.env.HICORTEX_REFLECT_MODEL ?? hcModel ?? "claude-sonnet-4-5-20250514",
+            reflectModel: process.env.HICORTEX_REFLECT_MODEL ?? hcModel ?? "claude-sonnet-4-6",
             provider,
         };
     }
@@ -101,7 +102,7 @@ function resolveLlmConfigForCC(overrides) {
             baseUrl: process.env.ANTHROPIC_BASE_URL ?? "https://api.anthropic.com",
             apiKey: anthropicKey,
             model: "claude-haiku-4-5-20251001",
-            reflectModel: "claude-sonnet-4-5-20250514",
+            reflectModel: "claude-sonnet-4-6",
             provider: "anthropic",
         };
     }
@@ -111,8 +112,8 @@ function resolveLlmConfigForCC(overrides) {
         return {
             baseUrl,
             apiKey: openaiKey,
-            model: "gpt-4o-mini",
-            reflectModel: "gpt-4o-mini",
+            model: "gpt-5.4-nano",
+            reflectModel: "gpt-5.4-nano",
             provider: detectProvider(baseUrl),
         };
     }
@@ -121,8 +122,8 @@ function resolveLlmConfigForCC(overrides) {
         return {
             baseUrl: "https://generativelanguage.googleapis.com/v1beta",
             apiKey: googleKey,
-            model: "gemini-2.0-flash",
-            reflectModel: "gemini-2.0-flash",
+            model: "gemini-2.5-flash",
+            reflectModel: "gemini-2.5-flash",
             provider: "google",
         };
     }
@@ -150,8 +151,6 @@ function detectProvider(url) {
         return "openrouter";
     if (u.includes("googleapis") || u.includes("generativelanguage"))
         return "google";
-    if (u.includes("z.ai") || u.includes("zai"))
-        return "zai";
     return "openai";
 }
 function readOpenClawConfig() {
@@ -162,7 +161,7 @@ function readOpenClawConfig() {
         const primary = config?.agents?.defaults?.model?.primary;
         if (!primary)
             return null;
-        // primary format is "provider/model" (e.g. "zai/glm-5-turbo", "openai/gpt-4o")
+        // primary format is "provider/model" (e.g. "openai/gpt-5", "anthropic/claude-sonnet-4-6")
         if (typeof primary === "string" && (primary.includes("/") || primary.includes(":"))) {
             const sep = primary.includes("/") ? "/" : ":";
             const [providerHint, ...rest] = primary.split(sep);
@@ -216,7 +215,7 @@ function readOcAuthKey(provider) {
                 const raw = (0, node_fs_1.readFileSync)(authPath, "utf-8");
                 const auth = JSON.parse(raw);
                 const profiles = auth?.profiles ?? {};
-                // Look for a profile matching the provider (e.g. "zai:default")
+                // Look for a profile matching the provider (e.g. "openai:default")
                 for (const [profileId, profile] of Object.entries(profiles)) {
                     const p = profile;
                     if (p?.provider === provider ||
@@ -245,8 +244,8 @@ function resolveFromEnv() {
         return {
             baseUrl,
             apiKey: openaiKey,
-            model: process.env.OPENAI_MODEL ?? "gpt-4o-mini",
-            reflectModel: process.env.OPENAI_MODEL ?? "gpt-4o-mini",
+            model: process.env.OPENAI_MODEL ?? "gpt-5.4-nano",
+            reflectModel: process.env.OPENAI_MODEL ?? "gpt-5.4-nano",
             provider,
         };
     }
@@ -255,8 +254,8 @@ function resolveFromEnv() {
         return {
             baseUrl: process.env.ANTHROPIC_BASE_URL ?? "https://api.anthropic.com",
             apiKey: anthropicKey,
-            model: "claude-sonnet-4-20250514",
-            reflectModel: "claude-sonnet-4-20250514",
+            model: "claude-sonnet-4-6",
+            reflectModel: "claude-sonnet-4-6",
             provider: "anthropic",
         };
     }
@@ -265,8 +264,8 @@ function resolveFromEnv() {
         return {
             baseUrl: "https://generativelanguage.googleapis.com/v1beta",
             apiKey: googleKey,
-            model: "gemini-2.0-flash",
-            reflectModel: "gemini-2.0-flash",
+            model: "gemini-2.5-flash",
+            reflectModel: "gemini-2.5-flash",
             provider: "google",
         };
     }
@@ -280,33 +279,23 @@ function getEnvKeyForProvider(provider) {
             return process.env.ANTHROPIC_API_KEY;
         case "google":
             return process.env.GOOGLE_API_KEY;
-        case "zai":
-            return process.env.ZAI_API_KEY ?? process.env.LLM_API_KEY;
         default:
             return undefined;
     }
 }
-/** Default base URLs for all OC-supported providers (from OC gateway binary). */
+/**
+ * Default base URLs for first-class supported providers.
+ *
+ * For any other provider, set llmBaseUrl explicitly in your config or use
+ * an OpenAI-compatible endpoint. The detectProvider() function will treat
+ * unknown URLs as openai-compatible by default.
+ */
 const PROVIDER_BASE_URLS = {
     openai: "https://api.openai.com/v1",
     anthropic: "https://api.anthropic.com",
     google: "https://generativelanguage.googleapis.com/v1beta",
     ollama: "http://localhost:11434",
     openrouter: "https://openrouter.ai/api",
-    zai: "https://api.z.ai/api/anthropic",
-    groq: "https://api.groq.com/openai/v1",
-    deepseek: "https://api.deepseek.com",
-    mistral: "https://api.mistral.ai/v1",
-    together: "https://api.together.xyz/v1",
-    perplexity: "https://api.perplexity.ai",
-    nvidia: "https://integrate.api.nvidia.com/v1",
-    xai: "https://api.x.ai/v1",
-    venice: "https://api.venice.ai/api/v1",
-    minimax: "https://api.minimaxi.com/v1",
-    moonshot: "https://api.moonshot.ai/v1",
-    kimi: "https://api.kimi.com/coding",
-    chutes: "https://api.chutes.ai",
-    kilo: "https://api.kilo.ai/api/gateway",
 };
 function getDefaultUrlForProvider(provider) {
     return PROVIDER_BASE_URLS[provider.toLowerCase()] ?? "https://api.openai.com/v1";
@@ -505,7 +494,7 @@ class LlmClient {
         if (this.config.provider === "ollama") {
             return this.completeOllama(model, prompt, maxTokens, timeoutMs);
         }
-        if (this.config.provider === "anthropic" || this.config.provider === "zai") {
+        if (this.config.provider === "anthropic") {
             return this.completeAnthropic(model, prompt, maxTokens, timeoutMs);
         }
         return this.completeOpenAiCompat(model, prompt, maxTokens, timeoutMs);
@@ -586,7 +575,7 @@ class LlmClient {
         return result.trim();
     }
     /**
-     * Anthropic Messages API (/v1/messages). Used for Anthropic and z.ai.
+     * Anthropic Messages API (/v1/messages).
      * Auth via x-api-key header.
      */
     async completeAnthropic(model, prompt, maxTokens, timeoutMs) {
@@ -622,8 +611,8 @@ class LlmClient {
      */
     async completeOpenAiCompat(model, prompt, maxTokens, timeoutMs) {
         const baseUrl = this.config.baseUrl.replace(/\/$/, "");
-        // Some providers (z.ai) include version in base URL already
-        const hasVersion = /\/v\d+\/?$/.test(baseUrl) || baseUrl.includes("/paas/v");
+        // Some providers include the API version in the base URL already
+        const hasVersion = /\/v\d+\/?$/.test(baseUrl);
         const url = hasVersion
             ? `${baseUrl}/chat/completions`
             : `${baseUrl}/v1/chat/completions`;
@@ -647,10 +636,6 @@ class LlmClient {
             this.handleRateLimit(resp);
         if (!resp.ok) {
             const text = await resp.text().catch(() => "");
-            // z.ai: "Insufficient balance" likely means wrong endpoint (coding vs paas)
-            if (text.includes("1113") || text.includes("Insufficient balance")) {
-                console.log(`[hicortex] LLM billing error. Check that llmBaseUrl matches your plan. Current: ${baseUrl}`);
-            }
             throw new Error(`LLM API error ${resp.status}: ${text}`);
         }
         const data = (await resp.json());

package/dist/mcp-server.js

@@ -55,7 +55,8 @@ const sse_js_1 = require("@modelcontextprotocol/sdk/server/sse.js");
 const zod_1 = require("zod");
 const db_js_1 = require("./db.js");
 const llm_js_1 = require("./llm.js");
-const license_js_1 = require("./license.js");
+const features_js_1 = require("./features.js");
+const state_js_1 = require("./state.js");
 const embedder_js_1 = require("./embedder.js");
 const storage = __importStar(require("./storage.js"));
 const retrieval = __importStar(require("./retrieval.js"));
@@ -121,12 +122,11 @@ function createMcpServer() {
     }, async ({ content, project, memory_type }) => {
         if (!db)
             return { content: [{ type: "text", text: "Hicortex not initialized" }], isError: true };
-        const features = (0, license_js_1.getFeatures)(stateDir);
-        if (features.maxMemories > 0 && storage.countMemories(db) >= features.maxMemories) {
+        if ((0, features_js_1.memoryCapReached)(storage.countMemories(db))) {
             return {
                 content: [{
                         type: "text",
-                        text: `Free tier limit reached (${features.maxMemories} memories). ` +
+                        text: `Free tier limit reached (${(0, features_js_1.maxMemoriesAllowed)()} memories). ` +
                             `Your existing memories and lessons still work — search and recall are unaffected. ` +
                             `New memories won't be saved until you upgrade.\n\n` +
                             `Upgrade for unlimited usage: https://hicortex.gamaze.com/`
@@ -296,11 +296,13 @@ async function startServer(options = {}) {
         ? `${llmConfig.reflectProvider}/${llmConfig.reflectModel}@${llmConfig.reflectBaseUrl}`
         : llmConfig.reflectModel;
     console.log(`[hicortex] LLM fast: ${llmConfig.provider}/${llmConfig.model}${distillInfo ? `, distill: ${distillInfo}` : ""}, reflect: ${reflectInfo}`);
-    // License: read from options, config file, or env var
+    // One-time migration of legacy state files (no-op if state.json exists)
+    (0, state_js_1.migrateLegacyState)(stateDir);
+    // License: read from options, config file, or env var, init feature cache
     const licenseKey = options.licenseKey
         ?? savedConfig?.licenseKey
         ?? process.env.HICORTEX_LICENSE_KEY;
-    (0, license_js_1.validateLicense)(licenseKey, stateDir).catch((err) => console.log(`[hicortex] License validation failed: ${err}`));
+    await (0, features_js_1.initFeatures)(licenseKey, stateDir);
     if (licenseKey) {
         console.log(`[hicortex] License key configured`);
     }
@@ -411,15 +413,12 @@ async function startServer(options = {}) {
         // Pro license blocks remote ingest (upgrade to Team for multi-client)
         const ip = req.ip ?? req.socket.remoteAddress ?? "";
         const isLocal = ip === "127.0.0.1" || ip === "::1" || ip === "::ffff:127.0.0.1";
-        if (!isLocal) {
-            const features = (0, license_js_1.getFeatures)(stateDir);
-            if (features.remoteIngest === false) {
-                res.status(403).json({
-                    error: "Pro license is single-machine. Upgrade to Team for multi-client remote ingestion.",
-                    upgrade: "https://hicortex.gamaze.com/",
-                });
-                return;
-            }
+        if (!isLocal && !(0, features_js_1.remoteIngestAllowed)()) {
+            res.status(403).json({
+                error: "Pro license is single-machine. Upgrade to Team for multi-client remote ingestion.",
+                upgrade: "https://hicortex.gamaze.com/",
+            });
+            return;
         }
         const { content, source_agent, project, memory_type, privacy, source_session, session_date } = req.body ?? {};
         if (!content || typeof content !== "string") {
@@ -440,9 +439,8 @@ async function startServer(options = {}) {
             }
         }
         // License check
-        const features = (0, license_js_1.getFeatures)(stateDir);
-        if (features.maxMemories > 0 && storage.countMemories(db) >= features.maxMemories) {
-            res.status(429).json({ error: "Memory limit reached", limit: features.maxMemories });
+        if ((0, features_js_1.memoryCapReached)(storage.countMemories(db))) {
+            res.status(429).json({ error: "Memory limit reached", limit: (0, features_js_1.maxMemoriesAllowed)() });
             return;
         }
         try {