feat: add skills uses to agent gateway

agent/src/agent.ts

@@ -1,10 +1,11 @@
 import { generateText, ModelMessage, stepCountIs, streamText, TextStreamPart, ToolSet } from 'ai'
 import { createChatGateway } from './gateway'
-import { ClientType, Schedule } from './types'
+import { AgentSkill, ClientType, Schedule } from './types'
 import { system, schedule } from './prompts'
 import { AuthFetcher } from './index'
 import { getScheduleTools } from './tools/schedule'
 import { getWebTools } from './tools/web'
+import { getSkillTools } from './tools/skill'
 
 export interface AgentParams {
   apiKey: string
@@ -19,6 +20,8 @@ export interface AgentParams {
   currentPlatform?: string
   braveApiKey?: string
   braveBaseUrl?: string
+  skills?: AgentSkill[]
+  useSkills?: string[]
 }
 
 export interface AgentInput {
@@ -28,6 +31,7 @@ export interface AgentInput {
 
 export interface AgentResult {
   messages: ModelMessage[]
+  skills: string[]
 }
 
 export const createAgent = (
@@ -36,13 +40,27 @@ export const createAgent = (
 ) => {
   const gateway = createChatGateway(params.clientType)
   const messages: ModelMessage[] = []
+  const enabledSkills: AgentSkill[] = params.skills ?? []
+  enabledSkills.push(
+    ...params.useSkills?.map((name) => params.skills?.find((s) => s.name === name)
+  ).filter((s) => s !== undefined) ?? [])
 
   const maxSteps = params.maxSteps ?? 50
 
   const getTools = () => {
     const scheduleTools = getScheduleTools({ fetch: fetcher })
+    const skillTools = getSkillTools({
+      skills: params.skills ?? [],
+      useSkill: (skill) => {
+        if (enabledSkills.some((s) => s.name === skill.name)) {
+          return
+        }
+        enabledSkills.push(skill)
+      }
+    })
     const tools: ToolSet = {
       ...scheduleTools,
+      ...skillTools,
     }
 
     // Add web search tools if Brave API key is provided
@@ -65,6 +83,8 @@ export const createAgent = (
       maxContextLoadTime: params.maxContextLoadTime,
       platforms: params.platforms ?? [],
       currentPlatform: params.currentPlatform,
+      skills: params.skills ?? [],
+      enabledSkills,
     })
   }
 
@@ -83,10 +103,16 @@ export const createAgent = (
       system: generateSystem(),
       stopWhen: stepCountIs(maxSteps),
       messages,
+      prepareStep: () => {
+        return {
+          system: generateSystem(),
+        }
+      },
       tools: getTools(),
     })
     return {
       messages: [user, ...response.messages],
+      skills: enabledSkills.map((s) => s.name),
     }
   }
 
@@ -105,6 +131,11 @@ export const createAgent = (
       system: generateSystem(),
       stopWhen: stepCountIs(maxSteps),
       messages,
+      prepareStep: () => {
+        return {
+          system: generateSystem(),
+        }
+      },
       tools: getTools(),
     })
     for await (const event of fullStream) {
@@ -112,6 +143,7 @@ export const createAgent = (
     }
     return {
       messages: [user, ...(await response).messages],
+      skills: enabledSkills.map((s) => s.name),
     }
   }
 
@@ -137,10 +169,16 @@ export const createAgent = (
       system: generateSystem(),
       stopWhen: stepCountIs(maxSteps),
       messages,
+      prepareStep: () => {
+        return {
+          system: generateSystem(),
+        }
+      },
       tools: getTools(),
     })
     return {
       messages: [user, ...response.messages],
+      skills: enabledSkills.map((s) => s.name),
     }
   }
 

agent/src/modules/chat.ts

@@ -7,6 +7,12 @@ import { ModelMessage } from 'ai'
 import { bearerMiddleware } from '../middlewares/bearer'
 import { loadConfig } from '../config'
 
+const Skill = z.object({
+  name: z.string().min(1, 'Skill name is required'),
+  description: z.string().min(1, 'Skill description is required'),
+  content: z.string().min(1, 'Skill content is required'),
+})
+
 const ChatBody = z.object({
   apiKey: z.string().min(1, 'API key is required'),
   baseUrl: z.string().min(1, 'Base URL is required'),
@@ -22,6 +28,8 @@ const ChatBody = z.object({
   maxContextLoadTime: z.number().min(1, 'Max context load time is required'),
   platforms: z.array(z.string()).optional(),
   currentPlatform: z.string().optional(),
+  skills: z.array(Skill).optional(),
+  useSkills: z.array(z.string()).optional(),
 
   messages: z.array(z.any()),
   query: z.string().min(1, 'Query is required'),
@@ -56,6 +64,8 @@ export const chatModule = new Elysia({ prefix: '/chat' })
       currentPlatform: body.currentPlatform,
       braveApiKey: config.brave?.api_key,
       braveBaseUrl: config.brave?.base_url,
+      skills: body.skills,
+      useSkills: body.useSkills,
     }, createAuthFetcher(bearer))
     try {
       const result = await ask({
@@ -98,6 +108,8 @@ export const chatModule = new Elysia({ prefix: '/chat' })
       currentPlatform: body.currentPlatform,
       braveApiKey: config.brave?.api_key,
       braveBaseUrl: config.brave?.base_url,
+      skills: body.skills,
+      useSkills: body.useSkills,
     }, createAuthFetcher(bearer))
     try {
       const streanGenerator = stream({
@@ -151,6 +163,8 @@ export const chatModule = new Elysia({ prefix: '/chat' })
       currentPlatform: body.currentPlatform,
       braveApiKey: config.brave?.api_key,
       braveBaseUrl: config.brave?.base_url,
+      skills: body.skills,
+      useSkills: body.useSkills,
     }, createAuthFetcher(bearer))
     try {
       return await triggerSchedule({

agent/src/prompts/system.ts

@@ -1,5 +1,6 @@
 import { time } from './shared'
 import { quote } from './utils'
+import { AgentSkill } from '../types'
 
 export interface SystemParams {
   date: Date
@@ -8,9 +9,20 @@ export interface SystemParams {
   maxContextLoadTime: number
   platforms: string[]
   currentPlatform?: string
+  skills: AgentSkill[]
+  enabledSkills: AgentSkill[]
 }
 
-export const system = ({ date, locale, language, maxContextLoadTime, platforms, currentPlatform }: SystemParams) => {
+export const skillPrompt = (skill: AgentSkill) => {
+  return `
+### ${skill.name}
+> ${skill.description}
+
+${skill.content}
+  `.trim()
+}
+
+export const system = ({ date, locale, language, maxContextLoadTime, platforms, currentPlatform, skills, enabledSkills }: SystemParams) => {
   return `
 ---
 ${time({ date, locale })}
@@ -46,5 +58,14 @@ Your abilities:
   + The ${quote('message')} is the message to send.
   + IF: the problem is initiated by a user, regardless of the platform the user is using, the content should be directly output in the content.
   + IF: the issue is initiated by a non-user (such as a scheduled task reminder), then it should be sent using the appropriate tools on the platform specified in the requirements.
+
+**Skills**
+
+There are ${skills.length} skills available, you can use ${quote('use_skill')} to use a skill.
+${skills.map(skill => `- ${skill.name}: ${skill.description}`).join('\n')}
+
+**Enabled Skills**
+
+${enabledSkills.map(skill => skillPrompt(skill)).join('\n\n---\n\n')}
   `.trim()
 }

agent/src/tools/skill.ts

@@ -0,0 +1,34 @@
+import { AgentSkill } from '../types'
+import { tool } from 'ai'
+import { z } from 'zod'
+
+interface SkillToolParams {
+  skills: AgentSkill[]
+  useSkill: (skill: AgentSkill, reason: string) => void
+}
+
+export const getSkillTools = ({ skills, useSkill }: SkillToolParams) => {
+  const useSkillTool = tool({
+    description: 'Use a skill if you think it is relevant to the current task',
+    inputSchema: z.object({
+      skillName: z.string().describe('The name of the skill to use'),
+      reason: z.string().describe('The reason why you think this skill is relevant to the current task'),
+    }),
+    execute: async ({ skillName, reason }) => {
+      const skill = skills.find((s) => s.name === skillName)
+      if (!skill) {
+        return { error: 'Skill not found' }
+      }
+      await useSkill(skill, reason)
+      return {
+        success: true,
+        skillName,
+        reason,
+      }
+    },
+  })
+
+  return {
+    'use_skill': useSkillTool,
+  }
+}

agent/src/types.ts

@@ -11,4 +11,10 @@ export interface Schedule {
   pattern: string
   maxCalls?: number
   command: string
+}
+
+export interface AgentSkill {
+  name: string
+  description: string
+  content: string
 }