opencode-swarm 3.2.1 → 3.3.1

package/dist/agents/sme/ai.d.ts

@@ -0,0 +1,2 @@
+import type { SMEDomainConfig } from './base';
+export declare const aiSMEConfig: SMEDomainConfig;

package/dist/config/constants.d.ts

@@ -1,9 +1,9 @@
-export declare const SME_AGENTS: readonly ["sme_windows", "sme_powershell", "sme_python", "sme_oracle", "sme_network", "sme_security", "sme_linux", "sme_vmware", "sme_azure", "sme_active_directory", "sme_ui_ux", "sme_web", "sme_database", "sme_devops", "sme_api"];
+export declare const SME_AGENTS: readonly ["sme_windows", "sme_powershell", "sme_python", "sme_oracle", "sme_network", "sme_security", "sme_linux", "sme_vmware", "sme_azure", "sme_active_directory", "sme_ui_ux", "sme_web", "sme_database", "sme_devops", "sme_api", "sme_ai"];
 export declare const QA_AGENTS: readonly ["security_reviewer", "auditor"];
 export declare const PIPELINE_AGENTS: readonly ["explorer", "coder", "test_engineer"];
 export declare const ORCHESTRATOR_NAME: "architect";
-export declare const ALL_SUBAGENT_NAMES: readonly ["sme_windows", "sme_powershell", "sme_python", "sme_oracle", "sme_network", "sme_security", "sme_linux", "sme_vmware", "sme_azure", "sme_active_directory", "sme_ui_ux", "sme_web", "sme_database", "sme_devops", "sme_api", "security_reviewer", "auditor", "explorer", "coder", "test_engineer"];
-export declare const ALL_AGENT_NAMES: readonly ["architect", "sme_windows", "sme_powershell", "sme_python", "sme_oracle", "sme_network", "sme_security", "sme_linux", "sme_vmware", "sme_azure", "sme_active_directory", "sme_ui_ux", "sme_web", "sme_database", "sme_devops", "sme_api", "security_reviewer", "auditor", "explorer", "coder", "test_engineer"];
+export declare const ALL_SUBAGENT_NAMES: readonly ["sme_windows", "sme_powershell", "sme_python", "sme_oracle", "sme_network", "sme_security", "sme_linux", "sme_vmware", "sme_azure", "sme_active_directory", "sme_ui_ux", "sme_web", "sme_database", "sme_devops", "sme_api", "sme_ai", "security_reviewer", "auditor", "explorer", "coder", "test_engineer"];
+export declare const ALL_AGENT_NAMES: readonly ["architect", "sme_windows", "sme_powershell", "sme_python", "sme_oracle", "sme_network", "sme_security", "sme_linux", "sme_vmware", "sme_azure", "sme_active_directory", "sme_ui_ux", "sme_web", "sme_database", "sme_devops", "sme_api", "sme_ai", "security_reviewer", "auditor", "explorer", "coder", "test_engineer"];
 export type SMEAgentName = (typeof SME_AGENTS)[number];
 export type QAAgentName = (typeof QA_AGENTS)[number];
 export type PipelineAgentName = (typeof PIPELINE_AGENTS)[number];

package/dist/config/schema.d.ts

@@ -29,6 +29,7 @@ export declare const PluginConfigSchema: z.ZodObject<{
         }, z.core.$strip>>>;
     }, z.core.$strip>>>;
     max_iterations: z.ZodDefault<z.ZodNumber>;
+    qa_retry_limit: z.ZodDefault<z.ZodNumber>;
     multi_domain_sme: z.ZodDefault<z.ZodBoolean>;
     auto_detect_domains: z.ZodDefault<z.ZodBoolean>;
     inject_phase_reminders: z.ZodDefault<z.ZodBoolean>;

package/dist/index.js

@@ -26,7 +26,8 @@ var SME_AGENTS = [
   "sme_web",
   "sme_database",
   "sme_devops",
-  "sme_api"
+  "sme_api",
+  "sme_ai"
 ];
 var QA_AGENTS = ["security_reviewer", "auditor"];
 var PIPELINE_AGENTS = ["explorer", "coder", "test_engineer"];
@@ -45,13 +46,13 @@ var CATEGORY_PREFIXES = {
   qa: "_qa"
 };
 var DEFAULT_MODELS = {
-  architect: "anthropic/claude-sonnet-4-5",
-  explorer: "google/gemini-2.0-flash",
-  coder: "anthropic/claude-sonnet-4-5",
-  test_engineer: "google/gemini-2.0-flash",
-  _sme: "google/gemini-2.0-flash",
-  _qa: "google/gemini-2.0-flash",
-  default: "google/gemini-2.0-flash"
+  architect: "kimi-for-coding/k2p5",
+  explorer: "opencode/gpt-5-nano",
+  coder: "zai-coding-plan/glm-4.7",
+  test_engineer: "zai-coding-plan/glm-4.7-flash",
+  _sme: "nvidia/openai/gpt-oss-120b",
+  _qa: "nvidia/openai/gpt-oss-120b",
+  default: "google/gemini-2.5-flash"
 };
 var DOMAIN_PATTERNS = {
   windows: [
@@ -13759,9 +13760,10 @@ var PluginConfigSchema = exports_external.object({
   agents: exports_external.record(exports_external.string(), AgentOverrideConfigSchema).optional(),
   swarms: exports_external.record(exports_external.string(), SwarmConfigSchema).optional(),
   max_iterations: exports_external.number().min(1).max(10).default(5),
+  qa_retry_limit: exports_external.number().min(1).max(10).default(3),
   multi_domain_sme: exports_external.boolean().default(true),
   auto_detect_domains: exports_external.boolean().default(true),
-  inject_phase_reminders: exports_external.boolean().default(false)
+  inject_phase_reminders: exports_external.boolean().default(true)
 });
 // src/config/loader.ts
 import * as fs from "fs";
@@ -13812,9 +13814,10 @@ function loadPluginConfig(directory) {
   const projectConfigPath = path.join(directory, ".opencode", CONFIG_FILENAME);
   let config2 = loadConfigFromPath(userConfigPath) ?? {
     max_iterations: 5,
+    qa_retry_limit: 3,
     multi_domain_sme: true,
     auto_detect_domains: true,
-    inject_phase_reminders: false
+    inject_phase_reminders: true
   };
   const projectConfig = loadConfigFromPath(projectConfigPath);
   if (projectConfig) {
@@ -13868,12 +13871,12 @@ You THINK. Subagents DO. You have the largest context window and strongest reaso
 1. DELEGATE all coding to @{{AGENT_PREFIX}}coder. You do NOT write code.
 2. ONE agent per message. Send, STOP, wait for response.
 3. ONE task per @{{AGENT_PREFIX}}coder call. Never batch.
-4. Fallback: Only code yourself after 3 @{{AGENT_PREFIX}}coder failures on same task.
+4. Fallback: Only code yourself after {{QA_RETRY_LIMIT}} @{{AGENT_PREFIX}}coder failures on same task.
 
 ## AGENTS
 
 @{{AGENT_PREFIX}}explorer - Codebase analysis
-@{{AGENT_PREFIX}}sme_[domain] - Domain expertise (windows, powershell, python, oracle, network, security, linux, vmware, azure, active_directory, ui_ux, web, database, devops, api)
+@{{AGENT_PREFIX}}sme_[domain] - Domain expertise (windows, powershell, python, oracle, network, security, linux, vmware, azure, active_directory, ui_ux, web, database, devops, api, ai)
 @{{AGENT_PREFIX}}coder - Implementation (one task at a time)
 @{{AGENT_PREFIX}}test_engineer - Test generation
 @{{AGENT_PREFIX}}security_reviewer - Vulnerability review
@@ -13969,8 +13972,8 @@ For each task (respecting dependencies):
 5c. @{{AGENT_PREFIX}}auditor - Verify
 5d. Result:
     - APPROVED \u2192 5e
-    - REJECTED (attempt 1-2) \u2192 Feedback to @{{AGENT_PREFIX}}coder, retry
-    - REJECTED (attempt 3) \u2192 Escalate, handle directly
+    - REJECTED (attempt < {{QA_RETRY_LIMIT}}) \u2192 Feedback to @{{AGENT_PREFIX}}coder, retry
+    - REJECTED (attempt {{QA_RETRY_LIMIT}}) \u2192 Escalate, handle directly
 5e. @{{AGENT_PREFIX}}test_engineer - Generate tests
 5f. Update plan.md [x], proceed to next task
 
@@ -14348,6 +14351,20 @@ var activeDirectorySMEConfig = {
 - ADSI/DirectoryServices .NET`
 };
 
+// src/agents/sme/ai.ts
+var aiSMEConfig = {
+  domain: "ai",
+  description: "AI/LLM systems and prompt engineering",
+  guidance: `- Prompt engineering (CoT, few-shot, structured output)
+- Context window management, token optimization
+- Model selection tradeoffs (cost, latency, capability)
+- Agent orchestration patterns (delegation, handoff)
+- RAG architectures, embedding strategies
+- Fine-tuning vs prompting decisions
+- Safety/alignment considerations
+- Tool use and function calling patterns`
+};
+
 // src/agents/sme/api.ts
 var apiSMEConfig = {
   domain: "api",
@@ -14561,7 +14578,8 @@ var SME_CONFIGS = {
   web: webSMEConfig,
   database: databaseSMEConfig,
   devops: devopsSMEConfig,
-  api: apiSMEConfig
+  api: apiSMEConfig,
+  ai: aiSMEConfig
 };
 var AGENT_TO_DOMAIN = {
   sme_windows: "windows",
@@ -14578,7 +14596,8 @@ var AGENT_TO_DOMAIN = {
   sme_web: "web",
   sme_database: "database",
   sme_devops: "devops",
-  sme_api: "api"
+  sme_api: "api",
+  sme_ai: "ai"
 };
 function createAllSMEAgents(getModel, loadPrompt) {
   return Object.entries(AGENT_TO_DOMAIN).map(([agentName, domain2]) => {
@@ -14633,11 +14652,12 @@ function applyOverrides(agent, swarmAgents, swarmPrefix) {
   }
   return agent;
 }
-function createSwarmAgents(swarmId, swarmConfig, isDefault) {
+function createSwarmAgents(swarmId, swarmConfig, isDefault, pluginConfig) {
   const agents = [];
   const swarmAgents = swarmConfig.agents;
   const prefix = isDefault ? "" : `${swarmId}_`;
   const swarmPrefix = isDefault ? undefined : swarmId;
+  const qaRetryLimit = pluginConfig?.qa_retry_limit ?? 3;
   const getModel = (baseName) => getModelForAgent(baseName, swarmAgents, swarmPrefix);
   const getPrompts = (name) => loadAgentPrompt(name);
   const prefixName = (name) => `${prefix}${name}`;
@@ -14649,7 +14669,7 @@ function createSwarmAgents(swarmId, swarmConfig, isDefault) {
     const swarmName = swarmConfig.name || swarmId;
     const swarmIdentity = isDefault ? "default" : swarmId;
     const agentPrefix = prefix;
-    architect.config.prompt = architect.config.prompt?.replace(/\{\{SWARM_ID\}\}/g, swarmIdentity).replace(/\{\{AGENT_PREFIX\}\}/g, agentPrefix);
+    architect.config.prompt = architect.config.prompt?.replace(/\{\{SWARM_ID\}\}/g, swarmIdentity).replace(/\{\{AGENT_PREFIX\}\}/g, agentPrefix).replace(/\{\{QA_RETRY_LIMIT\}\}/g, String(qaRetryLimit));
     if (!isDefault) {
       architect.description = `[${swarmName}] ${architect.description}`;
       const swarmHeader = `## \u26A0\uFE0F YOU ARE THE ${swarmName.toUpperCase()} SWARM ARCHITECT
@@ -14711,7 +14731,7 @@ function createAgents(config2) {
     for (const swarmId of swarmIds) {
       const swarmConfig = swarms[swarmId];
       const isDefault = swarmId === defaultSwarmId;
-      const swarmAgents = createSwarmAgents(swarmId, swarmConfig, isDefault);
+      const swarmAgents = createSwarmAgents(swarmId, swarmConfig, isDefault, config2);
       allAgents.push(...swarmAgents);
     }
   } else {
@@ -14719,7 +14739,7 @@ function createAgents(config2) {
       name: "Default",
       agents: config2?.agents
     };
-    const swarmAgents = createSwarmAgents("default", legacySwarmConfig, true);
+    const swarmAgents = createSwarmAgents("default", legacySwarmConfig, true, config2);
     allAgents.push(...swarmAgents);
   }
   return allAgents;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-swarm",
-  "version": "3.2.1",
+  "version": "3.3.1",
   "description": "Architect-centric agentic swarm plugin for OpenCode - hub-and-spoke orchestration with SME consultation, code generation, and QA review",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",