Azure model list

docs/src/content/docs/getting-started/configuration.mdx

@@ -450,6 +450,52 @@ script({
 
 </Steps>
 
+### Listing models
+
+In order to allow GenAIScript to list deployments in your Azure OpenAI service,
+you need to provide the Subscription ID **and you need to use Microsoft Entra!**.
+
+<Steps>
+
+<ol>
+
+<li>
+
+Open the Azure OpenAI resource in the [Azure Portal](https://portal.azure.com), open the **Overview** tab and copy the **Subscription ID**.
+
+</li>
+
+<li>
+
+Update the `.env` file with the subscription id.
+
+```txt title=".env"
+AZURE_OPENAI_SUBSCRIPTION_ID="..."
+```
+
+</li>
+
+<li>
+
+Test your configuration by running
+
+```sh
+npx genaiscript models azure
+```
+
+:::note
+
+This feature will probably not work with `AZURE_OPENAI_API_KEY`
+as the token does not have the proper scope to query the list of deployments.
+
+:::
+
+</li>
+
+</ol>
+
+</Steps>
+
 ### Custom credentials
 
 In some situations, the default credentials chain lookup may not work.

docs/src/content/docs/reference/cli/commands.md

@@ -634,7 +634,6 @@ Commands:
   help                      Show help for all commands
   system                    Show system information
   env [options] [provider]  Show .env information
-  models
 ```
 
 ### `info help`
@@ -673,23 +672,35 @@ Options:
   -h, --help    display help for command
 ```
 
-### `info models`
+## `models`
 
 ```
-Usage: genaiscript info models [options] [command]
+Usage: genaiscript models [options] [command]
 
 Options:
-  -h, --help      display help for command
+  -h, --help       display help for command
 
 Commands:
-  alias           Show model alias mapping
-  help [command]  display help for command
+  list [provider]  List all available models
+  alias            Show model alias mapping
+  help [command]   display help for command
+```
+
+### `models list`
+
+```
+Usage: genaiscript models list [options] [provider]
+
+List all available models
+
+Options:
+  -h, --help  display help for command
 ```
 
-#### `models alias`
+### `models alias`
 
 ```
-Usage: genaiscript info models alias [options]
+Usage: genaiscript models alias [options]
 
 Show model alias mapping
 

packages/cli/src/cli.ts

@@ -25,7 +25,13 @@ import {
 } from "./parse" // Parsing functions
 import { compileScript, createScript, fixScripts, listScripts } from "./scripts" // Script utilities
 import { codeQuery } from "./codequery" // Code parsing and query execution
-import { envInfo, modelAliasesInfo, scriptModelInfo, systemInfo } from "./info" // Information utilities
+import {
+    envInfo,
+    modelAliasesInfo,
+    modelList,
+    scriptModelInfo,
+    systemInfo,
+} from "./info" // Information utilities
 import { scriptTestList, scriptTestsView, scriptsTest } from "./test" // Test functions
 import { cacheClear } from "./cache" // Cache management
 import "node:console" // Importing console for side effects
@@ -528,7 +534,12 @@ export async function cli() {
         .option("-e, --error", "show errors")
         .option("-m, --models", "show models if possible")
         .action(envInfo) // Action to show environment information
-    const models = info.command("models")
+    const models = program.command("models")
+    models
+        .command("list", { isDefault: true })
+        .description("List all available models")
+        .arguments("[provider]")
+        .action(modelList)
     models
         .command("alias")
         .description("Show model alias mapping")

packages/cli/src/info.ts

@@ -15,7 +15,7 @@ import {
 import { CORE_VERSION } from "../../core/src/version"
 import { YAMLStringify } from "../../core/src/yaml"
 import { buildProject } from "./build"
-import { kMaxLength } from "buffer"
+import { deleteUndefinedValues } from "../../core/src/cleaners"
 
 /**
  * Outputs basic system information including node version, platform, architecture, and process ID.
@@ -106,3 +106,29 @@ export async function modelAliasesInfo() {
     )
     console.log(YAMLStringify(res))
 }
+
+/**
+ * Outputs environment information for model providers.
+ * @param provider - The specific provider to filter by (optional).
+ * @param options - Configuration options, including whether to show tokens.
+ */
+export async function modelList(
+    provider: string,
+    options?: { error?: boolean }
+) {
+    await runtimeHost.readConfig()
+    const providers = await resolveLanguageModelConfigurations(provider, {
+        ...(options || {}),
+        models: true,
+        error: true,
+    })
+    console.log(
+        YAMLStringify(
+            deleteUndefinedValues(
+                Object.fromEntries(
+                    providers.map((p) => [p.provider, p.error || p.models])
+                )
+            )
+        )
+    )
+}

packages/cli/src/nodehost.ts

@@ -24,6 +24,7 @@ import {
     AZURE_AI_INFERENCE_TOKEN_SCOPES,
     MODEL_PROVIDER_AZURE_SERVERLESS_OPENAI,
     DOT_ENV_FILENAME,
+    AZURE_MANAGEMENT_TOKEN_SCOPES,
 } from "../../core/src/constants"
 import { tryReadText } from "../../core/src/fs"
 import {
@@ -53,7 +54,7 @@ import {
     Project,
     ResponseStatus,
 } from "../../core/src/server/messages"
-import { createAzureTokenResolver } from "./azuretoken"
+import { createAzureTokenResolver } from "../../core/src/azuretoken"
 import {
     createAzureContentSafetyClient,
     isAzureContentSafetyClientConfigured,
@@ -98,13 +99,14 @@ export class NodeHost extends EventTarget implements RuntimeHost {
     readonly userInputQueue = new PLimitPromiseQueue(1)
     readonly azureToken: AzureTokenResolver
     readonly azureServerlessToken: AzureTokenResolver
+    readonly azureManagementToken: AzureTokenResolver
     readonly microsoftGraphToken: AzureTokenResolver
 
     constructor(dotEnvPath: string) {
         super()
         this.dotEnvPath = dotEnvPath
         this.azureToken = createAzureTokenResolver(
-            "Azure",
+            "Azure OpenAI",
             "AZURE_OPENAI_TOKEN_SCOPES",
             AZURE_COGNITIVE_SERVICES_TOKEN_SCOPES
         )
@@ -113,6 +115,11 @@ export class NodeHost extends EventTarget implements RuntimeHost {
             "AZURE_SERVERLESS_OPENAI_TOKEN_SCOPES",
             AZURE_AI_INFERENCE_TOKEN_SCOPES
         )
+        this.azureManagementToken = createAzureTokenResolver(
+            "Azure Management",
+            "AZURE_MANAGEMENT_TOKEN_SCOPES",
+            AZURE_MANAGEMENT_TOKEN_SCOPES
+        )
         this.microsoftGraphToken = createAzureTokenResolver(
             "Microsoft Graph",
             "MICROSOFT_GRAPH_TOKEN_SCOPES",

packages/core/src/azureopenai.ts

@@ -0,0 +1,98 @@
+import { LanguageModel, ListModelsFunction } from "./chat"
+import {
+    AZURE_MANAGEMENT_API_VERSION,
+    MODEL_PROVIDER_AZURE_OPENAI,
+} from "./constants"
+import { errorMessage, serializeError } from "./error"
+import { createFetch } from "./fetch"
+import { OpenAIChatCompletion } from "./openai"
+import { runtimeHost } from "./host"
+
+const listModels: ListModelsFunction = async (cfg, options) => {
+    try {
+        // Create a fetch instance to make HTTP requests
+        const { base } = cfg
+        const subscriptionId = process.env.AZURE_OPENAI_SUBSCRIPTION_ID
+        let resourceGroupName = process.env.AZURE_OPENAI_RESOURCE_GROUP
+        const accountName = /^https:\/\/([^\.]+)\./.exec(base)[1]
+
+        if (!subscriptionId || !accountName)
+            throw new Error("Missing subscriptionId, or accountName")
+        const token = await runtimeHost.azureManagementToken.token(
+            "default",
+            options
+        )
+        if (token.error) throw new Error(errorMessage(token.error))
+
+        const fetch = await createFetch({ retries: 0, ...options })
+        const get = async (url: string) => {
+            const res = await fetch(url, {
+                method: "GET",
+                headers: {
+                    Accept: "application/json",
+                    Authorization: `Bearer ${token.token.token}`,
+                },
+            })
+            if (res.status !== 200)
+                return {
+                    ok: false,
+                    status: res.status,
+                    error: serializeError(res.statusText),
+                }
+            return await res.json()
+        }
+
+        if (!resourceGroupName) {
+            const resources: {
+                value: {
+                    id: string
+                    name: string
+                    type: "OpenAI"
+                }[]
+            } = await get(
+                `https://management.azure.com/subscriptions/${subscriptionId}/resources?api-version=2021-04-01`
+            )
+            const resource = resources.value.find((r) => r.name === accountName)
+            resourceGroupName = /\/resourceGroups\/([^\/]+)\/providers\//.exec(
+                resource?.id
+            )[1]
+            if (!resourceGroupName) throw new Error("Resource group not found")
+        }
+
+        // https://learn.microsoft.com/en-us/rest/api/aiservices/accountmanagement/deployments/list-skus?view=rest-aiservices-accountmanagement-2024-10-01&tabs=HTTP
+        const deployments: {
+            value: {
+                id: string
+                name: string
+                properties: {
+                    model: {
+                        format: string
+                        name: string
+                        version: string
+                    }
+                }
+            }[]
+        } = await get(
+            `https://management.azure.com/subscriptions/${subscriptionId}/resourceGroups/${resourceGroupName}/providers/Microsoft.CognitiveServices/accounts/${accountName}/deployments/?api-version=${AZURE_MANAGEMENT_API_VERSION}`
+        )
+        return {
+            ok: true,
+            models: deployments.value.map((model) => ({
+                id: model.name,
+                family: model.properties.model.name,
+                details: `${model.properties.model.format} ${model.properties.model.name}`,
+                url: `https://ai.azure.com/resource/deployments/${encodeURIComponent(model.id)}`,
+                version: model.properties.model.version,
+            })),
+        }
+    } catch (e) {
+        return { ok: false, error: serializeError(e) }
+    }
+}
+
+// Define the Ollama model with its completion handler and model listing function
+export const AzureOpenAIModel = Object.freeze<LanguageModel>({
+    id: MODEL_PROVIDER_AZURE_OPENAI,
+    completer: OpenAIChatCompletion,
+    listModels,
+})

packages/cli/src/azuretoken.ts → packages/core/src/azuretoken.ts

@@ -119,7 +119,6 @@ class AzureTokenResolverImpl implements AzureTokenResolver {
         if (this._token || this._error)
             return { token: this._token, error: this._error }
         if (!this._resolver) {
-            logVerbose(`${this.name}: creating token`)
             const scope = await runtimeHost.readSecret(this.envName)
             const scopes = scope ? scope.split(",") : this.scopes
             this._resolver = createAzureToken(

packages/core/src/constants.ts

@@ -15,13 +15,17 @@ export const MAX_TOOL_CALLS = 10000
 // https://github.com/Azure/azure-rest-api-specs/blob/main/specification/cognitiveservices/data-plane/AzureOpenAI/inference/stable/2024-02-01/inference.yaml
 // https://learn.microsoft.com/en-us/azure/ai-services/openai/api-version-deprecation
 export const AZURE_OPENAI_API_VERSION = "2024-06-01"
+export const AZURE_MANAGEMENT_API_VERSION = "2024-10-01"
 export const AZURE_COGNITIVE_SERVICES_TOKEN_SCOPES = Object.freeze([
     "https://cognitiveservices.azure.com/.default",
 ])
 export const AZURE_AI_INFERENCE_VERSION = "2024-08-01-preview"
 export const AZURE_AI_INFERENCE_TOKEN_SCOPES = Object.freeze([
     "https://ml.azure.com/.default",
 ])
+export const AZURE_MANAGEMENT_TOKEN_SCOPES = Object.freeze([
+    "https://management.azure.com/.default",
+])
 export const AZURE_TOKEN_EXPIRATION = 59 * 60_000 // 59 minutes
 
 export const TOOL_URL = "https://microsoft.github.io/genaiscript"

packages/core/src/host.ts

@@ -134,9 +134,10 @@ export interface RuntimeHost extends Host {
     project: Project
     workspace: Omit<WorkspaceFileSystem, "grep" | "writeCached">
 
-    azureToken: AzureTokenResolver
-    azureServerlessToken: AzureTokenResolver
-    microsoftGraphToken: AzureTokenResolver
+    azureToken?: AzureTokenResolver
+    azureServerlessToken?: AzureTokenResolver
+    azureManagementToken?: AzureTokenResolver
+    microsoftGraphToken?: AzureTokenResolver
 
     modelAliases: Readonly<ModelConfigurations>
     clientLanguageModel?: LanguageModel

packages/core/src/llms.json

@@ -51,6 +51,12 @@
                     "description": "Azure OpenAI API key. **You do NOT need this if you are using Microsoft Entra ID.",
                     "secret": true
                 },
+                "AZURE_OPENAI_SUBSCRIPTION_ID": {
+                    "description": "Azure OpenAI subscription ID to list available deployments (Microsoft Entra only)."
+                },
+                "AZURE_OPENAI_API_VERSION": {
+                    "description": "Azure OpenAI API version."
+                },
                 "AZURE_OPENAI_API_CREDENTIALS": {
                     "description": "Azure OpenAI API credentials type. Leave as 'default' unless you have a special Azure setup.",
                     "enum": [

packages/core/src/lm.ts

@@ -12,6 +12,7 @@ import {
     MODEL_PROVIDER_TRANSFORMERS,
     MODEL_PROVIDERS,
     MODEL_WHISPERASR_PROVIDER,
+    MODEL_PROVIDER_AZURE_OPENAI,
 } from "./constants"
 import { runtimeHost } from "./host"
 import { OllamaModel } from "./ollama"
@@ -20,13 +21,15 @@ import { TransformersModel } from "./transformers"
 import { GitHubModel } from "./github"
 import { LMStudioModel } from "./lmstudio"
 import { WhiserAsrModel } from "./whisperasr"
+import { AzureOpenAIModel } from "./azureopenai"
 
 export function resolveLanguageModel(provider: string): LanguageModel {
     if (provider === MODEL_PROVIDER_GITHUB_COPILOT_CHAT) {
         const m = runtimeHost.clientLanguageModel
         if (!m) throw new Error("Github Copilot Chat Models not available")
         return m
     }
+    if (provider === MODEL_PROVIDER_AZURE_OPENAI) return AzureOpenAIModel
     if (provider === MODEL_PROVIDER_GITHUB) return GitHubModel
     if (provider === MODEL_PROVIDER_OLLAMA) return OllamaModel
     if (provider === MODEL_PROVIDER_AICI) return AICIModel