@johpaz/hive-sdk 0.4.5 → 0.4.6

package/README.md

@@ -271,4 +271,4 @@ npm view @johpaz/hive-sdk dist-tags   # verificar después del release
 
 ---
 
-*Hive SDK v0.4.5 — MIT*
+*Hive SDK v0.4.6 — MIT*

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@johpaz/hive-sdk",
-  "version": "0.4.5",
+  "version": "0.4.6",
   "private": false,
   "description": "Hive SDK — The Agent Harness SDK. Build, deploy, and scale AI agent applications with multi-channel support, context engineering, and swarm orchestration.",
   "license": "MIT",
@@ -99,7 +99,7 @@
     "drift": "bun scripts/check-drift.ts"
   },
   "dependencies": {
-    "@johpaz/hive-db": "^0.4.0",
+    "@johpaz/hive-db": "^0.5.1",
     "@anthropic-ai/sdk": "^0.74.0",
     "@google/genai": "^1.43.0",
     "@modelcontextprotocol/sdk": "^1.26.0",

package/packages/core/src/agent/agent-loop.ts

@@ -521,6 +521,7 @@ export async function* runAgent(
         messages: clearOldToolResults(messages) as LLMMessage[],
         tools: ctx.tools.length > 0 ? ctx.tools : undefined,
         signal: opts.signal,
+        sessionId: opts.threadId,
         onToken: opts.onToken && !delegationGroupAtCall
           ? (token: string) => {
             streamedThisCall = true
@@ -1133,6 +1134,7 @@ export async function* runAgent(
         ...providerCfg,
         messages: clearOldToolResults(messages) as LLMMessage[],
         tools: undefined, // no tools — force text response
+        sessionId: opts.threadId,
       })
       if (synthesis.usage) {
         totalInputTokens += synthesis.usage.input_tokens

package/packages/core/src/agent/capability-search.ts

@@ -81,7 +81,8 @@ export async function searchCapabilities(
   const k = opts.k ?? 10;
   const types = opts.types?.length ? opts.types : undefined;
   const trimmed = query.trim();
-  if (!trimmed) return [];
+  // hive-db >= 0.4 rejects k <= 0 instead of returning no hits.
+  if (!trimmed || k <= 0) return [];
 
   const startTime = performance.now();
   const db = await getHiveDb();

package/packages/core/src/agent/llm-client.ts

@@ -97,6 +97,13 @@ export interface LLMCallOptions {
   signal?: AbortSignal
   /** Enable extended thinking for supported models (Anthropic Claude 3.7+). */
   thinking?: { enabled: boolean; budget_tokens?: number }
+  /**
+   * Stable id of the conversation this call belongs to (the agent loop's
+   * threadId). Providers that route or cache per session derive their own
+   * header from it — OpenCode Go's `x-opencode-session` — and never send it
+   * verbatim, since it carries user, channel and peer ids.
+   */
+  sessionId?: string
 }
 
 export interface LLMResponse {
@@ -214,6 +221,14 @@ export function describeProviderFailure(
   provider: string,
   cleanModel: string,
 ): string {
+  // NVIDIA usa 404 también para modelos que siguen en su catálogo público pero
+  // no están habilitados para esa key ("Function '…': Not found for account
+  // '…'"). Decir que "se retiró" mandaba a buscar un modelo que sí existe.
+  if (status === 404 && /not found for account/i.test((err as Error)?.message ?? "")) {
+    return `Tu cuenta de ${provider} no tiene habilitado el modelo "${cleanModel}" (HTTP 404): `
+      + `figura en su catálogo, pero no está disponible para esta API key. `
+      + `Elige otro modelo en Ajustes → Proveedores.`
+  }
   if (status === 404 || status === 410) {
     return `El modelo "${cleanModel}" ya no existe en ${provider} (HTTP ${status}). `
       + `El proveedor lo retiró de su catálogo; reintentar no sirve. `

package/packages/core/src/agent/llm-providers/openai-compat-base.ts

@@ -82,8 +82,8 @@ export abstract class OpenAICompatBase implements LLMProvider {
   /** Override to true for providers running on localhost. */
   protected isLocalProvider(): boolean { return false }
 
-  /** Override to customize the OpenAI client (e.g. strip unwanted headers, add custom fetch). */
-  protected async resolveOpenAIClient(apiKey: string, baseURL: string | undefined): Promise<any> {
+  /** Override to customize the OpenAI client (e.g. strip unwanted headers, add custom fetch, per-call headers). */
+  protected async resolveOpenAIClient(apiKey: string, baseURL: string | undefined, _options?: LLMCallOptions): Promise<any> {
     const { default: OpenAI } = await import("openai")
     return new OpenAI({ apiKey, baseURL })
   }
@@ -135,7 +135,7 @@ export abstract class OpenAICompatBase implements LLMProvider {
       throw new Error(`API key missing for provider: ${this.providerName}. Configure it in Settings → Providers.`)
     }
 
-    const client = await this.resolveOpenAIClient(apiKey, baseURL)
+    const client = await this.resolveOpenAIClient(apiKey, baseURL, options)
 
     const sanitized = sanitizeMessages(options.messages)
     const rawMessages = this.needsReasoningRoundtrip()

package/packages/core/src/agent/llm-providers/opencode-go.ts

@@ -1,4 +1,29 @@
+import { createHash, randomUUID } from "node:crypto"
+import pkg from "../../../../../package.json"
 import { OpenAICompatBase } from "./openai-compat-base.ts"
+import type { LLMCallOptions } from "../llm-client.ts"
+
+/**
+ * OpenCode Go rechaza con 400 `MissingSessionID` toda petición sin
+ * `x-opencode-session` (verificado 2026-09-10 con una key real: sin el header
+ * 400, con él 200). Su documentación pide además que el cliente se identifique
+ * con su propio User-Agent en vez del genérico del SDK:
+ * https://opencode.ai/docs/go/#where-can-i-use-it
+ */
+const USER_AGENT = `hive-sdk/${pkg.version}`
+
+/** Sesión de respaldo para llamadas sin conversación (compactación, metas). */
+const PROCESS_SESSION = randomUUID()
+
+/**
+ * La sesión tiene que ser estable por conversación —OpenCode la usa para rutear
+ * y cachear el prompt—, pero el threadId lleva usuario, canal y contacto: a
+ * OpenCode sólo le llega un hash.
+ */
+function sessionHeader(sessionId: string | undefined): string {
+  if (!sessionId) return PROCESS_SESSION
+  return createHash("sha256").update(`hive:${sessionId}`).digest("hex").slice(0, 32)
+}
 
 export class OpenCodeGoProvider extends OpenAICompatBase {
   static readonly secretKey = "OPENCODE_GO_API_KEY"
@@ -6,4 +31,16 @@ export class OpenCodeGoProvider extends OpenAICompatBase {
   constructor() {
     super("opencode-go")
   }
+
+  protected async resolveOpenAIClient(apiKey: string, baseURL: string | undefined, options?: LLMCallOptions): Promise<any> {
+    const { default: OpenAI } = await import("openai")
+    return new OpenAI({
+      apiKey,
+      baseURL,
+      defaultHeaders: {
+        "x-opencode-session": sessionHeader(options?.sessionId),
+        "User-Agent": USER_AGENT,
+      },
+    })
+  }
 }

package/packages/core/src/storage/seed.ts

@@ -266,7 +266,9 @@ export const SEED_DATA: SeedData = {
     { id: "z-ai/glm-5.3-flash", providerId: "openrouter", name: "GLM 5.3 Flash (OR)", modelType: "llm", contextWindow: 1310720, capabilities: JSON.stringify(["chat", "vision", "json_mode", "function_calling", "streaming", "code", "reasoning"]), inputPer1M: 0.075, outputPer1M: 0.25 },
     { id: "z-ai/glm-5.2", providerId: "openrouter", name: "GLM 5.2 (OR)", modelType: "llm", contextWindow: 1048576, capabilities: JSON.stringify(["chat", "json_mode", "function_calling", "streaming", "code", "reasoning"]), inputPer1M: 0.966, outputPer1M: 3.036 },
     // Qwen
-    { id: "qwen/qwen3.8-max", providerId: "openrouter", name: "Qwen3.8 Max (OR)", modelType: "llm", contextWindow: 1000000, capabilities: JSON.stringify(["chat", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 2, outputPer1M: 6 },
+    // OpenRouter renombró qwen/qwen3.8-max a su versión fechada: mismo contexto
+    // y precio, ahora también con imagen. Verificado 2026-09-10.
+    { id: "qwen/qwen3.8-max-0902", providerId: "openrouter", name: "Qwen3.8 Max (OR)", modelType: "llm", contextWindow: 1000000, capabilities: JSON.stringify(["chat", "vision", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 2, outputPer1M: 6 },
     { id: "qwen/qwen3.8-flash", providerId: "openrouter", name: "Qwen3.8 Flash (OR)", modelType: "llm", contextWindow: 1000000, capabilities: JSON.stringify(["chat", "vision", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 0.15, outputPer1M: 0.47 },
     { id: "qwen/qwen3.7-flash", providerId: "openrouter", name: "Qwen3.7 Flash (OR)", modelType: "llm", contextWindow: 1000000, capabilities: JSON.stringify(["chat", "json_mode", "function_calling", "streaming"]), inputPer1M: 0.03, outputPer1M: 0.13 },
     // xAI
@@ -334,8 +336,10 @@ export const SEED_DATA: SeedData = {
     // provider `z-ai` directo, o el enrutado de OpenRouter.
     { id: "moonshotai/kimi-k3", providerId: "nvidia", name: "Kimi K3 (NVIDIA)", modelType: "llm", contextWindow: 262144, capabilities: JSON.stringify(["chat", "code", "vision", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 0, outputPer1M: 0 },
     { id: "nvidia/nemotron-3.5-lightning-30b-a3b", providerId: "nvidia", name: "Nemotron 3.5 Lightning 30B", modelType: "llm", contextWindow: 262144, capabilities: JSON.stringify(["chat", "code", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 0, outputPer1M: 0 },
-    { id: "moonshotai/kimi-k2.6", providerId: "nvidia", name: "Kimi K2.6 (NVIDIA)", modelType: "llm", contextWindow: 262144, capabilities: JSON.stringify(["chat", "code", "vision", "function_calling", "streaming", "reasoning"]), inputPer1M: 0, outputPer1M: 0 },
-    { id: "minimaxai/minimax-m3", providerId: "nvidia", name: "MiniMax M3 (NVIDIA)", modelType: "llm", contextWindow: 1000000, capabilities: JSON.stringify(["chat", "code", "vision", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 0, outputPer1M: 0 },
+    // MiniMax M3 (minimaxai/minimax-m3) se sacó: NVIDIA lo retiró (410 Gone, ya
+    // no figura en /v1/models). Kimi K2.6 (moonshotai/kimi-k2.6) se sacó por lo
+    // mismo que DeepSeek V4 Pro: 404 "Function not found for account" en dos
+    // cuentas distintas. Verificado 2026-09-10.
     { id: "nvidia/nemotron-3-ultra-550b-a55b", providerId: "nvidia", name: "Nemotron 3 Ultra 550B", modelType: "llm", contextWindow: 1000000, capabilities: JSON.stringify(["chat", "code", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 0, outputPer1M: 0 },
     { id: "nvidia/nemotron-3-super-120b-a12b", providerId: "nvidia", name: "Nemotron 3 Super 120B", modelType: "llm", contextWindow: 1000000, capabilities: JSON.stringify(["chat", "code", "json_mode", "function_calling", "streaming", "reasoning"]), inputPer1M: 0, outputPer1M: 0 },
     // DeepSeek V4 Pro (deepseek-ai/deepseek-v4-pro) se sacó: devuelve 404