Merge branch 'master' into pdf-screen-template

lerna.json

@@ -1,6 +1,6 @@
 {
   "$schema": "node_modules/lerna/schemas/lerna-schema.json",
-  "version": "3.8.0",
+  "version": "3.8.1",
   "npmClient": "yarn",
   "concurrency": 20,
   "command": {

packages/backend-core/src/cache/writethrough.ts

@@ -96,6 +96,24 @@ async function get<T extends Document>(db: Database, id: string): Promise<T> {
   return cacheItem.doc
 }
 
+async function tryGet<T extends Document>(
+  db: Database,
+  id: string
+): Promise<T | null> {
+  const cache = await getCache()
+  const cacheKey = makeCacheKey(db, id)
+  let cacheItem: CacheItem<T> | null = await cache.get(cacheKey)
+  if (!cacheItem) {
+    const doc = await db.tryGet<T>(id)
+    if (!doc) {
+      return null
+    }
+    cacheItem = makeCacheItem(doc)
+    await cache.store(cacheKey, cacheItem)
+  }
+  return cacheItem.doc
+}
+
 async function remove(db: Database, docOrId: any, rev?: any): Promise<void> {
   const cache = await getCache()
   if (!docOrId) {
@@ -123,10 +141,17 @@ export class Writethrough {
     return put(this.db, doc, writeRateMs)
   }
 
+  /**
+   * @deprecated use `tryGet` instead
+   */
   async get<T extends Document>(id: string) {
     return get<T>(this.db, id)
   }
 
+  async tryGet<T extends Document>(id: string) {
+    return tryGet<T>(this.db, id)
+  }
+
   async remove(docOrId: any, rev?: any) {
     return remove(this.db, docOrId, rev)
   }

packages/backend-core/src/configs/configs.ts

@@ -47,6 +47,9 @@ export async function getConfig<T extends Config>(
 export async function save(
   config: Config
 ): Promise<{ id: string; rev: string }> {
+  if (!config._id) {
+    config._id = generateConfigID(config.type)
+  }
   const db = context.getGlobalDB()
   return db.put(config)
 }

packages/backend-core/src/configs/tests/configs.spec.ts

@@ -12,7 +12,6 @@ describe("configs", () => {
 
   const setDbPlatformUrl = async (dbUrl: string) => {
     const settingsConfig = {
-      _id: configs.generateConfigID(ConfigType.SETTINGS),
       type: ConfigType.SETTINGS,
       config: {
         platformUrl: dbUrl,

packages/backend-core/src/constants/db.ts

@@ -60,6 +60,11 @@ export const StaticDatabases = {
   SCIM_LOGS: {
     name: "scim-logs",
   },
+  // Used by self-host users making use of Budicloud resources. Introduced when
+  // we started letting self-host users use Budibase AI in the cloud.
+  SELF_HOST_CLOUD: {
+    name: "self-host-cloud",
+  },
 }
 
 export const APP_PREFIX = prefixed(DocumentType.APP)

packages/backend-core/src/context/mainContext.ts

@@ -157,6 +157,33 @@ export async function doInTenant<T>(
   return newContext(updates, task)
 }
 
+// We allow self-host licensed users to make use of some Budicloud services
+// (e.g. Budibase AI). When they do this, they use their license key as an API
+// key. We use that license key to identify the tenant ID, and we set the
+// context to be self-host using cloud. This affects things like where their
+// quota documents get stored (because we want to avoid creating a new global
+// DB for each self-host tenant).
+export async function doInSelfHostTenantUsingCloud<T>(
+  tenantId: string,
+  task: () => T
+): Promise<T> {
+  const updates = { tenantId, isSelfHostUsingCloud: true }
+  return newContext(updates, task)
+}
+
+export function isSelfHostUsingCloud() {
+  const context = Context.get()
+  return !!context?.isSelfHostUsingCloud
+}
+
+export function getSelfHostCloudDB() {
+  const context = Context.get()
+  if (!context || !context.isSelfHostUsingCloud) {
+    throw new Error("Self-host cloud DB not found")
+  }
+  return getDB(StaticDatabases.SELF_HOST_CLOUD.name)
+}
+
 export async function doInAppContext<T>(
   appId: string,
   task: () => T
@@ -325,6 +352,11 @@ export function getGlobalDB(): Database {
   if (!context || (env.MULTI_TENANCY && !context.tenantId)) {
     throw new Error("Global DB not found")
   }
+  if (context.isSelfHostUsingCloud) {
+    throw new Error(
+      "Global DB not found - self-host users using cloud don't have a global DB"
+    )
+  }
   return getDB(baseGlobalDBName(context?.tenantId))
 }
 
@@ -344,6 +376,11 @@ export function getAppDB(opts?: any): Database {
   if (!appId) {
     throw new Error("Unable to retrieve app DB - no app ID.")
   }
+  if (isSelfHostUsingCloud()) {
+    throw new Error(
+      "App DB not found - self-host users using cloud don't have app DBs"
+    )
+  }
   return getDB(appId, opts)
 }
 

packages/backend-core/src/context/types.ts

@@ -5,6 +5,7 @@ import { GoogleSpreadsheet } from "google-spreadsheet"
 // keep this out of Budibase types, don't want to expose context info
 export type ContextMap = {
   tenantId?: string
+  isSelfHostUsingCloud?: boolean
   appId?: string
   identity?: IdentityContext
   environmentVariables?: Record<string, string>

packages/pro

@@ -1 +1 @@
-Subproject commit 40c36f86584568d31abd6dd5b6b00dd3a458093f
+Subproject commit 8eb981cf01151261697a8f26c08c4c28f66b8e15

packages/server/package.json

@@ -49,6 +49,7 @@
   "author": "Budibase",
   "license": "GPL-3.0",
   "dependencies": {
+    "@anthropic-ai/sdk": "^0.27.3",
     "@apidevtools/swagger-parser": "10.0.3",
     "@aws-sdk/client-dynamodb": "3.709.0",
     "@aws-sdk/client-s3": "3.709.0",

packages/server/scripts/dev/manage.js

@@ -47,6 +47,7 @@ async function init() {
     VERSION: "0.0.0+local",
     PASSWORD_MIN_LENGTH: "1",
     OPENAI_API_KEY: "sk-abcdefghijklmnopqrstuvwxyz1234567890abcd",
+    BUDICLOUD_URL: "https://budibaseqa.app",
   }
 
   config = { ...config, ...existingConfig }

packages/server/src/api/routes/tests/ai.spec.ts

@@ -0,0 +1,344 @@
+import { mockChatGPTResponse } from "../../../tests/utilities/mocks/ai/openai"
+import TestConfiguration from "../../../tests/utilities/TestConfiguration"
+import nock from "nock"
+import { configs, env, features, setEnv } from "@budibase/backend-core"
+import {
+  AIInnerConfig,
+  ConfigType,
+  License,
+  PlanModel,
+  PlanType,
+  ProviderConfig,
+} from "@budibase/types"
+import { context } from "@budibase/backend-core"
+import { mocks } from "@budibase/backend-core/tests"
+import { MockLLMResponseFn } from "../../../tests/utilities/mocks/ai"
+import { mockAnthropicResponse } from "../../../tests/utilities/mocks/ai/anthropic"
+
+function dedent(str: string) {
+  return str
+    .split("\n")
+    .map(line => line.trim())
+    .join("\n")
+}
+
+type SetupFn = (
+  config: TestConfiguration
+) => Promise<() => Promise<void> | void>
+interface TestSetup {
+  name: string
+  setup: SetupFn
+  mockLLMResponse: MockLLMResponseFn
+}
+
+function budibaseAI(): SetupFn {
+  return async () => {
+    const cleanup = setEnv({
+      OPENAI_API_KEY: "test-key",
+    })
+    mocks.licenses.useBudibaseAI()
+    return async () => {
+      mocks.licenses.useCloudFree()
+      cleanup()
+    }
+  }
+}
+
+function customAIConfig(providerConfig: Partial<ProviderConfig>): SetupFn {
+  return async (config: TestConfiguration) => {
+    mocks.licenses.useAICustomConfigs()
+
+    const innerConfig: AIInnerConfig = {
+      myaiconfig: {
+        provider: "OpenAI",
+        name: "OpenAI",
+        apiKey: "test-key",
+        defaultModel: "gpt-4o-mini",
+        active: true,
+        isDefault: true,
+        ...providerConfig,
+      },
+    }
+
+    const { id, rev } = await config.doInTenant(
+      async () =>
+        await configs.save({
+          type: ConfigType.AI,
+          config: innerConfig,
+        })
+    )
+
+    return async () => {
+      mocks.licenses.useCloudFree()
+
+      await config.doInTenant(async () => {
+        const db = context.getGlobalDB()
+        await db.remove(id, rev)
+      })
+    }
+  }
+}
+
+const allProviders: TestSetup[] = [
+  {
+    name: "OpenAI API key",
+    setup: async () => {
+      return setEnv({
+        OPENAI_API_KEY: "test-key",
+      })
+    },
+    mockLLMResponse: mockChatGPTResponse,
+  },
+  {
+    name: "OpenAI API key with custom config",
+    setup: customAIConfig({ provider: "OpenAI", defaultModel: "gpt-4o-mini" }),
+    mockLLMResponse: mockChatGPTResponse,
+  },
+  {
+    name: "Anthropic API key with custom config",
+    setup: customAIConfig({
+      provider: "Anthropic",
+      defaultModel: "claude-3-5-sonnet-20240620",
+    }),
+    mockLLMResponse: mockAnthropicResponse,
+  },
+  {
+    name: "BudibaseAI",
+    setup: budibaseAI(),
+    mockLLMResponse: mockChatGPTResponse,
+  },
+]
+
+describe("AI", () => {
+  const config = new TestConfiguration()
+
+  beforeAll(async () => {
+    await config.init()
+  })
+
+  afterAll(() => {
+    config.end()
+  })
+
+  beforeEach(() => {
+    nock.cleanAll()
+  })
+
+  describe.each(allProviders)(
+    "provider: $name",
+    ({ setup, mockLLMResponse }: TestSetup) => {
+      let cleanup: () => Promise<void> | void
+      beforeAll(async () => {
+        cleanup = await setup(config)
+      })
+
+      afterAll(async () => {
+        const maybePromise = cleanup()
+        if (maybePromise) {
+          await maybePromise
+        }
+      })
+
+      describe("POST /api/ai/js", () => {
+        let cleanup: () => void
+        beforeAll(() => {
+          cleanup = features.testutils.setFeatureFlags("*", {
+            AI_JS_GENERATION: true,
+          })
+        })
+
+        afterAll(() => {
+          cleanup()
+        })
+
+        it("handles correct plain code response", async () => {
+          mockLLMResponse(`return 42`)
+
+          const { code } = await config.api.ai.generateJs({ prompt: "test" })
+          expect(code).toBe("return 42")
+        })
+
+        it("handles correct markdown code response", async () => {
+          mockLLMResponse(
+            dedent(`
+                \`\`\`js
+                return 42
+                \`\`\`
+            `)
+          )
+
+          const { code } = await config.api.ai.generateJs({ prompt: "test" })
+          expect(code).toBe("return 42")
+        })
+
+        it("handles multiple markdown code blocks returned", async () => {
+          mockLLMResponse(
+            dedent(`
+                This:
+
+                \`\`\`js
+                return 42
+                \`\`\`
+
+                Or this:
+
+                \`\`\`js
+                return 10
+                \`\`\`
+            `)
+          )
+
+          const { code } = await config.api.ai.generateJs({ prompt: "test" })
+          expect(code).toBe("return 42")
+        })
+
+        // TODO: handle when this happens
+        it.skip("handles no code response", async () => {
+          mockLLMResponse("I'm sorry, you're quite right, etc.")
+          const { code } = await config.api.ai.generateJs({ prompt: "test" })
+          expect(code).toBe("")
+        })
+
+        it("handles LLM errors", async () => {
+          mockLLMResponse(() => {
+            throw new Error("LLM error")
+          })
+          await config.api.ai.generateJs({ prompt: "test" }, { status: 500 })
+        })
+      })
+
+      describe("POST /api/ai/cron", () => {
+        it("handles correct cron response", async () => {
+          mockLLMResponse("0 0 * * *")
+
+          const { message } = await config.api.ai.generateCron({
+            prompt: "test",
+          })
+          expect(message).toBe("0 0 * * *")
+        })
+
+        it("handles expected LLM error", async () => {
+          mockLLMResponse("Error generating cron: skill issue")
+
+          await config.api.ai.generateCron(
+            {
+              prompt: "test",
+            },
+            { status: 400 }
+          )
+        })
+
+        it("handles unexpected LLM error", async () => {
+          mockLLMResponse(() => {
+            throw new Error("LLM error")
+          })
+
+          await config.api.ai.generateCron(
+            {
+              prompt: "test",
+            },
+            { status: 500 }
+          )
+        })
+      })
+    }
+  )
+})
+
+describe("BudibaseAI", () => {
+  const config = new TestConfiguration()
+  let cleanup: () => void | Promise<void>
+  beforeAll(async () => {
+    await config.init()
+    cleanup = await budibaseAI()(config)
+  })
+
+  afterAll(async () => {
+    if ("then" in cleanup) {
+      await cleanup()
+    } else {
+      cleanup()
+    }
+    config.end()
+  })
+
+  describe("POST /api/ai/chat", () => {
+    let envCleanup: () => void
+    let featureCleanup: () => void
+    beforeAll(() => {
+      envCleanup = setEnv({ SELF_HOSTED: false })
+      featureCleanup = features.testutils.setFeatureFlags("*", {
+        AI_JS_GENERATION: true,
+      })
+    })
+
+    afterAll(() => {
+      featureCleanup()
+      envCleanup()
+    })
+
+    beforeEach(() => {
+      nock.cleanAll()
+      const license: License = {
+        plan: {
+          type: PlanType.FREE,
+          model: PlanModel.PER_USER,
+          usesInvoicing: false,
+        },
+        features: [],
+        quotas: {} as any,
+        tenantId: config.tenantId,
+      }
+      nock(env.ACCOUNT_PORTAL_URL).get("/api/license").reply(200, license)
+    })
+
+    it("handles correct chat response", async () => {
+      mockChatGPTResponse("Hi there!")
+      const { message } = await config.api.ai.chat({
+        messages: [{ role: "user", content: "Hello!" }],
+        licenseKey: "test-key",
+      })
+      expect(message).toBe("Hi there!")
+    })
+
+    it("handles chat response error", async () => {
+      mockChatGPTResponse(() => {
+        throw new Error("LLM error")
+      })
+      await config.api.ai.chat(
+        {
+          messages: [{ role: "user", content: "Hello!" }],
+          licenseKey: "test-key",
+        },
+        { status: 500 }
+      )
+    })
+
+    it("handles no license", async () => {
+      nock.cleanAll()
+      nock(env.ACCOUNT_PORTAL_URL).get("/api/license").reply(404)
+      await config.api.ai.chat(
+        {
+          messages: [{ role: "user", content: "Hello!" }],
+          licenseKey: "test-key",
+        },
+        {
+          status: 403,
+        }
+      )
+    })
+
+    it("handles no license key", async () => {
+      await config.api.ai.chat(
+        {
+          messages: [{ role: "user", content: "Hello!" }],
+          // @ts-expect-error - intentionally wrong
+          licenseKey: undefined,
+        },
+        {
+          status: 403,
+        }
+      )
+    })
+  })
+})

packages/server/src/api/routes/tests/row.spec.ts

@@ -46,7 +46,7 @@ import { withEnv } from "../../../environment"
 import { JsTimeoutError } from "@budibase/string-templates"
 import { isDate } from "../../../utilities"
 import nock from "nock"
-import { mockChatGPTResponse } from "../../../tests/utilities/mocks/openai"
+import { mockChatGPTResponse } from "../../../tests/utilities/mocks/ai/openai"
 
 const timestamp = new Date("2023-01-26T11:48:57.597Z").toISOString()
 tk.freeze(timestamp)

packages/server/src/api/routes/tests/search.spec.ts

@@ -44,7 +44,7 @@ import { generator, structures, mocks } from "@budibase/backend-core/tests"
 import { DEFAULT_EMPLOYEE_TABLE_SCHEMA } from "../../../db/defaultData/datasource_bb_default"
 import { generateRowIdField } from "../../../integrations/utils"
 import { cloneDeep } from "lodash/fp"
-import { mockChatGPTResponse } from "../../../tests/utilities/mocks/openai"
+import { mockChatGPTResponse } from "../../../tests/utilities/mocks/ai/openai"
 
 const descriptions = datasourceDescribe({ plus: true })
 

packages/server/src/api/routes/tests/viewV2.spec.ts

@@ -41,7 +41,7 @@ import { datasourceDescribe } from "../../../integrations/tests/utils"
 import merge from "lodash/merge"
 import { quotas } from "@budibase/pro"
 import { context, db, events, roles, setEnv } from "@budibase/backend-core"
-import { mockChatGPTResponse } from "../../../tests/utilities/mocks/openai"
+import { mockChatGPTResponse } from "../../../tests/utilities/mocks/ai/openai"
 import nock from "nock"
 
 const descriptions = datasourceDescribe({ plus: true })

packages/server/src/automations/tests/steps/openai.spec.ts

@@ -1,11 +1,8 @@
 import { createAutomationBuilder } from "../utilities/AutomationTestBuilder"
-import { setEnv as setCoreEnv } from "@budibase/backend-core"
+import { setEnv as setCoreEnv, withEnv } from "@budibase/backend-core"
 import { Model, MonthlyQuotaName, QuotaUsageType } from "@budibase/types"
 import TestConfiguration from "../../..//tests/utilities/TestConfiguration"
-import {
-  mockChatGPTError,
-  mockChatGPTResponse,
-} from "../../../tests/utilities/mocks/openai"
+import { mockChatGPTResponse } from "../../../tests/utilities/mocks/ai/openai"
 import nock from "nock"
 import { mocks } from "@budibase/backend-core/tests"
 import { quotas } from "@budibase/pro"
@@ -83,7 +80,9 @@ describe("test the openai action", () => {
   })
 
   it("should present the correct error message when an error is thrown from the createChatCompletion call", async () => {
-    mockChatGPTError()
+    mockChatGPTResponse(() => {
+      throw new Error("oh no")
+    })
 
     const result = await expectAIUsage(0, () =>
       createAutomationBuilder(config)
@@ -108,11 +107,13 @@ describe("test the openai action", () => {
     // path, because we've enabled Budibase AI. The exact value depends on a
     // calculation we use to approximate cost. This uses Budibase's OpenAI API
     // key, so we charge users for it.
-    const result = await expectAIUsage(14, () =>
-      createAutomationBuilder(config)
-        .onAppAction()
-        .openai({ model: Model.GPT_4O_MINI, prompt: "Hello, world" })
-        .test({ fields: {} })
+    const result = await withEnv({ SELF_HOSTED: false }, () =>
+      expectAIUsage(14, () =>
+        createAutomationBuilder(config)
+          .onAppAction()
+          .openai({ model: Model.GPT_4O_MINI, prompt: "Hello, world" })
+          .test({ fields: {} })
+      )
     )
 
     expect(result.steps[0].outputs.response).toEqual("This is a test")

packages/server/src/constants/screens.ts

@@ -365,7 +365,11 @@ export function createSampleDataTableScreen(): Screen {
           _component: "@budibase/standard-components/textv2",
           _styles: {
             normal: {
+              "--grid-desktop-col-start": 1,
               "--grid-desktop-col-end": 3,
+              "--grid-desktop-row-start": 1,
+              "--grid-desktop-row-end": 3,
+              "--grid-mobile-col-end": 7,
             },
             hover: {},
             active: {},
@@ -384,6 +388,7 @@ export function createSampleDataTableScreen(): Screen {
               "--grid-desktop-row-start": 1,
               "--grid-desktop-row-end": 3,
               "--grid-desktop-h-align": "end",
+              "--grid-mobile-col-start": 7,
             },
             hover: {},
             active: {},

packages/server/src/tests/utilities/api/ai.ts

@@ -0,0 +1,47 @@
+import {
+  ChatCompletionRequest,
+  ChatCompletionResponse,
+  GenerateCronRequest,
+  GenerateCronResponse,
+  GenerateJsRequest,
+  GenerateJsResponse,
+} from "@budibase/types"
+import { Expectations, TestAPI } from "./base"
+import { constants } from "@budibase/backend-core"
+
+export class AIAPI extends TestAPI {
+  generateJs = async (
+    req: GenerateJsRequest,
+    expectations?: Expectations
+  ): Promise<GenerateJsResponse> => {
+    return await this._post<GenerateJsResponse>(`/api/ai/js`, {
+      body: req,
+      expectations,
+    })
+  }
+
+  generateCron = async (
+    req: GenerateCronRequest,
+    expectations?: Expectations
+  ): Promise<GenerateCronResponse> => {
+    return await this._post<GenerateCronResponse>(`/api/ai/cron`, {
+      body: req,
+      expectations,
+    })
+  }
+
+  chat = async (
+    req: ChatCompletionRequest & { licenseKey: string },
+    expectations?: Expectations
+  ): Promise<ChatCompletionResponse> => {
+    const headers: Record<string, string> = {}
+    if (req.licenseKey) {
+      headers[constants.Header.LICENSE_KEY] = req.licenseKey
+    }
+    return await this._post<ChatCompletionResponse>(`/api/ai/chat`, {
+      body: req,
+      headers,
+      expectations,
+    })
+  }
+}

packages/server/src/tests/utilities/api/index.ts

@@ -22,8 +22,10 @@ import { UserPublicAPI } from "./public/user"
 import { MiscAPI } from "./misc"
 import { OAuth2API } from "./oauth2"
 import { AssetsAPI } from "./assets"
+import { AIAPI } from "./ai"
 
 export default class API {
+  ai: AIAPI
   application: ApplicationAPI
   attachment: AttachmentAPI
   automation: AutomationAPI
@@ -52,6 +54,7 @@ export default class API {
   }
 
   constructor(config: TestConfiguration) {
+    this.ai = new AIAPI(config)
     this.application = new ApplicationAPI(config)
     this.attachment = new AttachmentAPI(config)
     this.automation = new AutomationAPI(config)

packages/server/src/tests/utilities/mocks/ai/anthropic.ts

@@ -0,0 +1,48 @@
+import AnthropicClient from "@anthropic-ai/sdk"
+import nock from "nock"
+import { MockLLMResponseFn, MockLLMResponseOpts } from "."
+
+let chatID = 1
+const SPACE_REGEX = /\s+/g
+
+export const mockAnthropicResponse: MockLLMResponseFn = (
+  answer: string | ((prompt: string) => string),
+  opts?: MockLLMResponseOpts
+) => {
+  return nock(opts?.host || "https://api.anthropic.com")
+    .post("/v1/messages")
+    .reply((uri: string, body: nock.Body) => {
+      const req = body as AnthropicClient.MessageCreateParamsNonStreaming
+      const prompt = req.messages[0].content
+      if (typeof prompt !== "string") {
+        throw new Error("Anthropic mock only supports string prompts")
+      }
+
+      let content
+      if (typeof answer === "function") {
+        try {
+          content = answer(prompt)
+        } catch (e) {
+          return [500, "Internal Server Error"]
+        }
+      } else {
+        content = answer
+      }
+
+      const resp: AnthropicClient.Messages.Message = {
+        id: `${chatID++}`,
+        type: "message",
+        role: "assistant",
+        model: req.model,
+        stop_reason: "end_turn",
+        usage: {
+          input_tokens: prompt.split(SPACE_REGEX).length,
+          output_tokens: content.split(SPACE_REGEX).length,
+        },
+        stop_sequence: null,
+        content: [{ type: "text", text: content }],
+      }
+      return [200, resp]
+    })
+    .persist()
+}

packages/server/src/tests/utilities/mocks/ai/index.ts

@@ -0,0 +1,10 @@
+import { Scope } from "nock"
+
+export interface MockLLMResponseOpts {
+  host?: string
+}
+
+export type MockLLMResponseFn = (
+  answer: string | ((prompt: string) => string),
+  opts?: MockLLMResponseOpts
+) => Scope

packages/server/src/tests/utilities/mocks/ai/openai.ts

@@ -1,12 +1,9 @@
 import nock from "nock"
+import { MockLLMResponseFn, MockLLMResponseOpts } from "."
 
 let chatID = 1
 const SPACE_REGEX = /\s+/g
 
-interface MockChatGPTResponseOpts {
-  host?: string
-}
-
 interface Message {
   role: string
   content: string
@@ -47,19 +44,24 @@ interface ChatCompletionResponse {
   usage: Usage
 }
 
-export function mockChatGPTResponse(
+export const mockChatGPTResponse: MockLLMResponseFn = (
   answer: string | ((prompt: string) => string),
-  opts?: MockChatGPTResponseOpts
-) {
+  opts?: MockLLMResponseOpts
+) => {
   return nock(opts?.host || "https://api.openai.com")
     .post("/v1/chat/completions")
-    .reply(200, (uri: string, requestBody: ChatCompletionRequest) => {
-      const messages = requestBody.messages
+    .reply((uri: string, body: nock.Body) => {
+      const req = body as ChatCompletionRequest
+      const messages = req.messages
       const prompt = messages[0].content
 
       let content
       if (typeof answer === "function") {
-        content = answer(prompt)
+        try {
+          content = answer(prompt)
+        } catch (e) {
+          return [500, "Internal Server Error"]
+        }
       } else {
         content = answer
       }
@@ -76,7 +78,7 @@ export function mockChatGPTResponse(
         id: `chatcmpl-${chatID}`,
         object: "chat.completion",
         created: Math.floor(Date.now() / 1000),
-        model: requestBody.model,
+        model: req.model,
         system_fingerprint: `fp_${chatID}`,
         choices: [
           {
@@ -97,14 +99,7 @@ export function mockChatGPTResponse(
           },
         },
       }
-      return response
+      return [200, response]
     })
     .persist()
 }
-
-export function mockChatGPTError() {
-  return nock("https://api.openai.com")
-    .post("/v1/chat/completions")
-    .reply(500, "Internal Server Error")
-    .persist()
-}

packages/types/src/api/web/ai.ts

@@ -1,5 +1,18 @@
 import { EnrichedBinding } from "../../ui"
 
+export interface Message {
+  role: "system" | "user"
+  content: string
+}
+
+export interface ChatCompletionRequest {
+  messages: Message[]
+}
+
+export interface ChatCompletionResponse {
+  message?: string
+}
+
 export interface GenerateJsRequest {
   prompt: string
   bindings?: EnrichedBinding[]
@@ -8,3 +21,11 @@ export interface GenerateJsRequest {
 export interface GenerateJsResponse {
   code: string
 }
+
+export interface GenerateCronRequest {
+  prompt: string
+}
+
+export interface GenerateCronResponse {
+  message?: string
+}

packages/types/src/documents/global/config.ts

@@ -117,6 +117,7 @@ export type AIProvider =
   | "AzureOpenAI"
   | "TogetherAI"
   | "Custom"
+  | "BudibaseAI"
 
 export interface ProviderConfig {
   provider: AIProvider