feat: Add Cloudflare AI Provider (#817)

README.md

@@ -361,6 +361,17 @@ Here are some of the available scripts:
 - `db:generate`: Generate new db migrations (and create the db if it doesn't already exist). Note that after generating a new migration, you'll have to `npm i` to copy the migrations into `dist/`.
 - `db:migrate`: Run existing db migrations (and create the db if it doesn't already exist)
 
+To run the CLI during development you can run a command like: `npm run local -- eval --config $(readlink -f ./examples/cloudflare-ai/chat_config.yaml)`, where any parts of the command after `--` are passed through to our CLI entrypoint. Since the Next dev server isn't supported in this mode, see the instructions above for running the web server.
+
 # [» View full documentation «](https://promptfoo.dev/docs/intro)
 
 [providers-docs]: https://promptfoo.dev/docs/providers
+
+
+### Adding a New Provider
+
+1. Create an implementation in `src/providers/SOME_PROVIDER_FILE`
+2. Update `loadApiProvider` in `src/providers.ts` to load your provider via string
+3. Add test cases in `test/providers.test.ts`
+   1. Test the actual provider implementation
+   2. Test loading the provider via a `loadApiProvider` test

examples/cloudflare-ai/README.md

@@ -0,0 +1,5 @@
+# Cloudflare Workers AI
+
+See the [full docs for Workers AI](https://developers.cloudflare.com/workers-ai/) and the [Cloudflare REST docs](https://developers.cloudflare.com/api/operations/workers-ai-post-run-model) for further reference on endpoints.
+
+The basic example in this repo shows the difference in outputs between the "chat" and "completion" versions of the provider.

examples/cloudflare-ai/chat_advanced_configuration.yaml

@@ -0,0 +1,25 @@
+prompts:
+  - Tell me a really funny joke about {{topic}}. The joke should contain the word {{topic}}
+
+providers:
+  - id: cloudflare-ai:chat:@cf/meta/llama-3-8b-instruct
+    config:
+      accountId: YOUR_ACCOUNT_ID_HERE
+      # ===============
+      # It is not recommended to keep your API key on the config file since it is a secret value.
+      # Use the CLOUDFLARE_API_KEY environment variable or set the apiKeyEnvar value
+      # in the config
+      # apiKey: YOUR_API_KEY_HERE
+      # apiKeyEnvar: SOME_ENV_HAR_CONTAINING_THE_API_KEY
+      # ===============
+      # Additional model parameters that are passed through in the HTTP request body to the Cloudflare REST API call
+      # to run the model
+      max_tokens: 800
+      seed: 1
+
+tests:
+  - vars:
+      topic: birds
+    assert:
+      - type: icontains
+        value: "{{topic}}"

examples/cloudflare-ai/chat_config.yaml

@@ -0,0 +1,18 @@
+prompts:
+  - What is the capital of the {{country}}?
+
+providers:
+  - cloudflare-ai:completion:@cf/meta/llama-3-8b-instruct
+  - cloudflare-ai:chat:@cf/meta/llama-3-8b-instruct
+
+tests:
+  - vars:
+      country: United States
+    assert:
+      - type: icontains
+        value: Washington, D.C.
+  - vars:
+      country: England
+    assert:
+      - type: icontains
+        value: London

examples/cloudflare-ai/embedding_configuration.yaml

@@ -0,0 +1,35 @@
+prompts:
+  - This prompt will be evaluated for embedding similarity
+
+defaultTest:
+  options:
+    provider:
+      embedding:
+        id: cloudflare-ai:embedding:@cf/baai/bge-base-en-v1.5
+        config:
+          # accountId: YOUR_ACCOUNT_ID_HERE
+
+
+
+providers:
+  - id: cloudflare-ai:chat:@cf/meta/llama-3-8b-instruct
+    config:
+      # accountId: YOUR_ACCOUNT_ID_HERE
+
+      # ===============
+      # It is not recommended to keep your API key on the config file since it is a secret value.
+      # Use the CLOUDFLARE_API_KEY environment variable or set the apiKeyEnvar value
+      # in the config
+      # apiKey: YOUR_API_KEY_HERE
+      # apiKeyEnvar: SOME_ENV_HAR_CONTAINING_THE_API_KEY
+      # ===============
+      # Additional model parameters that are passed through in the HTTP request body to the Cloudflare REST API call
+      # to run the model
+      # max_tokens: 800
+      # seed: 1
+
+tests:
+  - assert:
+      - type: similar
+        value: embedding similarity
+        threshold: 0.6

jest.config.ts

@@ -1,14 +1,12 @@
 /** @type {import('ts-jest/dist/types').InitialOptionsTsJest} */
 import type { Config } from 'jest';
+import type { TsJestTransformerOptions } from 'ts-jest';
+
+const tsJestConfig: TsJestTransformerOptions & Record<string, unknown> = { useESM: true };
 
 const config: Config = {
   transform: {
-    '\\.[jt]sx?$': 'ts-jest',
-  },
-  globals: {
-    'ts-jest': {
-      useESM: true,
-    },
+    '^.+\\.m?[tj]sx?$': ['ts-jest', tsJestConfig],
   },
   /*
   moduleNameMapper: {

package.json

@@ -25,8 +25,9 @@
     "promptfoo": "dist/src/main.js"
   },
   "scripts": {
+    "bin": "dist/src/main.js",
     "tsc": "tsc",
-    "local": "ts-node --esm --files src/main.ts",
+    "local": "ts-node --cwdMode --transpileOnly src/main.ts",
     "local:web": "cd src/web/nextui && npm run dev",
     "install:nextui": "cd src/web/nextui && npm install",
     "build:clean": "rm -rf dist",
@@ -71,6 +72,7 @@
     "@types/semver": "^7.5.0",
     "@types/uuid": "^9.0.2",
     "babel-jest": "^29.5.0",
+    "cloudflare": "^3.2.0",
     "drizzle-kit": "^0.20.13",
     "jest": "^29.5.0",
     "jest-watch-typeahead": "^2.2.2",

site/docs/providers/cloudflare-ai.md

@@ -0,0 +1,53 @@
+# Cloudflare Workers AI
+
+This provider supports the [models](https://developers.cloudflare.com/workers-ai/models/) provided by Cloudflare Workers AI, a serverless edge embedding and inference runtime.
+
+## Required Configuration
+
+Calling the Workers AI requires the user to supply a Cloudflare account ID and API key with sufficient permissions to invoke the Workers AI REST endpoints.
+
+```bash
+export CLOUDFLARE_ACCOUNT_ID=YOUR_ACCOUNT_ID_HERE
+export CLOUDFLARE_API_KEY=YOUR_API_KEY_HERE
+```
+
+The Cloudflare account ID is not secret and therefore it is safe to put it in your `promptfoo` configuration file. The Cloudflare API key is secret, so while you can provide it in the config, this is **HIGHLY NOT RECOMMENDED** as it might lead to abuse. See below for an example safe configuration:
+
+
+```yaml
+prompts:
+  - Tell me a really funny joke about {{topic}}. The joke should contain the word {{topic}}
+
+providers:
+  - id: cloudflare-ai:chat:@cf/meta/llama-3-8b-instruct
+    config:
+      accountId: YOUR_ACCOUNT_ID_HERE
+      # It is not recommended to keep your API key on the config file since it is a secret value.
+      # Use the CLOUDFLARE_API_KEY environment variable or set the apiKeyEnvar value
+      # in the config
+      # apiKey: YOUR_API_KEY_HERE
+      # apiKeyEnvar: SOME_ENV_HAR_CONTAINING_THE_API_KEY
+
+tests:
+  - vars:
+      topic: birds
+    assert:
+      - type: icontains
+        value: "{{topic}}"
+```
+
+In addition to `apiKeyEnvar` allowed environment variable redirection for the `CLOUDFLARE_API_KEY` value, the `accountIdEnvar` can be used to similarly redirect to a value for the `CLOUDFLARE_ACCOUNT_ID`.
+
+## Available Models and Model Parameters
+
+Cloudflare is constantly adding new models to its inventory. See their [official list of models](https://developers.cloudflare.com/workers-ai/models/) for a list of supported models. Different models support different parameters, which is supported by supplying those parameters as additional keys of the config object in the `promptfoo` config file.
+
+For an example of how advanced embedding configuration should be supplied, see `examples/cloudflare-ai/embedding_configuration.yaml`
+
+For an example of how advanced completion/chat configuration should be supplied, see `examples/cloudflare-ai/chat_advanced_configuration.yaml`
+
+Different models support different parameters. While this provider strives to be relatively flexible, it is possible that not all possible parameters have been added. If you need support for a parameter that is not yet supported, please open up a PR to add support.
+
+## Future Improvements
+
+- [ ] Allow for the pass through of all generic configuration parameters for Cloudflare REST API

src/main.ts

@@ -781,6 +781,12 @@ async function main() {
             const rowsLeft = summary.table.body.length - 25;
             logger.info(`... ${rowsLeft} more row${rowsLeft === 1 ? '' : 's'} not shown ...\n`);
           }
+        } else if (summary.stats.failures !== 0) {
+          logger.debug(
+            `At least one evaluation failure occurred. This might be caused by the underlying call to the provider, or a test failure. Context: \n${JSON.stringify(
+              summary.results,
+            )}`,
+          );
         }
 
         const { outputPath } = config;

src/providers.ts

@@ -47,6 +47,7 @@ import {
 import { AwsBedrockCompletionProvider, AwsBedrockEmbeddingProvider } from './providers/bedrock';
 import { PythonProvider } from './providers/pythonCompletion';
 import { CohereChatCompletionProvider } from './providers/cohere';
+import * as CloudflareAiProviders from './providers/cloudflare-ai';
 import { BAMChatProvider, BAMEmbeddingProvider } from './providers/bam';
 import { PortkeyChatCompletionProvider } from './providers/portkey';
 import { HttpProvider } from './providers/http';
@@ -289,6 +290,32 @@ export async function loadApiProvider(
         `Invalid BAM provider: ${providerPath}. Use one of the following providers: bam:chat:<model name>`,
       );
     }
+  } else if (providerPath.startsWith('cloudflare-ai:')) {
+    // Load Cloudflare AI
+    const splits = providerPath.split(':');
+    const modelType = splits[1];
+    const deploymentName = splits[2];
+
+    if (modelType === 'chat') {
+      ret = new CloudflareAiProviders.CloudflareAiChatCompletionProvider(
+        deploymentName,
+        providerOptions,
+      );
+    } else if (modelType === 'embedding' || modelType === 'embeddings') {
+      ret = new CloudflareAiProviders.CloudflareAiEmbeddingProvider(
+        deploymentName,
+        providerOptions,
+      );
+    } else if (modelType === 'completion') {
+      ret = new CloudflareAiProviders.CloudflareAiCompletionProvider(
+        deploymentName,
+        providerOptions,
+      );
+    } else {
+      throw new Error(
+        `Unknown Cloudflare AI model type: ${modelType}. Use one of the following providers: cloudflare-ai:chat:<model name>, cloudflare-ai:completion:<model name>, cloudflare-ai:embedding:`,
+      );
+    }
   } else if (providerPath.startsWith('webhook:')) {
     const webhookUrl = providerPath.substring('webhook:'.length);
     ret = new WebhookProvider(webhookUrl, providerOptions);