feat: migrate cloudflare client to openai-compatible

Argcfile.sh

@@ -81,6 +81,7 @@ test-server() {
 OPENAI_COMPATIBLE_PLATFORMS=( \
   openai,gpt-4o-mini,https://api.openai.com/v1 \
   ai21,jamba-1.5-mini,https://api.ai21.com/studio/v1 \
+  cloudflare,@cf/meta/llama-3.1-8b-instruct, \
   deepinfra,meta-llama/Meta-Llama-3.1-8B-Instruct,https://api.deepinfra.com/v1/openai \
   deepseek,deepseek-chat,https://api.deepseek.com \
   fireworks,accounts/fireworks/models/llama-v3p1-8b-instruct,https://api.fireworks.ai/inference/v1 \
@@ -111,7 +112,7 @@ chat() {
     fi
     for platform_config in "${OPENAI_COMPATIBLE_PLATFORMS[@]}"; do
         if [[ "$argc_platform" == "${platform_config%%,*}" ]]; then
-            api_base="${platform_config##*,}"
+            _retrieve_api_base
             break
         fi
     done
@@ -141,15 +142,15 @@ chat() {
 models() {
     for platform_config in "${OPENAI_COMPATIBLE_PLATFORMS[@]}"; do
         if [[ "$argc_platform" == "${platform_config%%,*}" ]]; then
-            api_base="${platform_config##*,}"
+            _retrieve_api_base
             break
         fi
     done
     if [[ -n "$api_base" ]]; then
         env_prefix="$(echo "$argc_platform" | tr '[:lower:]' '[:upper:]')"
         api_key_env="${env_prefix}_API_KEY"
         api_key="${!api_key_env}" 
-        _openai_models
+        _retrieve_models
     else
         argc models-$argc_platform
     fi
@@ -173,7 +174,7 @@ chat-openai-compatible() {
 # @option --api-base! $$
 # @option --api-key! $$
 models-openai-compatible() {
-    _openai_models
+    _retrieve_models
 }
 
 # @cmd Chat with azure-openai api
@@ -271,19 +272,6 @@ chat-vertexai() {
 -d "$(_build_body vertexai "$@")" 
 }
 
-# @cmd Chat with cloudflare api
-# @env CLOUDFLARE_API_KEY!
-# @option -m --model=@cf/meta/llama-3-8b-instruct $CLOUDFLARE_MODEL
-# @flag -S --no-stream
-# @arg text~
-chat-cloudflare() {
-    url="https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run/$argc_model"
-    _wrapper curl -i "$url" \
--X POST \
--H "Authorization: Bearer $CLOUDFLARE_API_KEY" \
--d "$(_build_body cloudflare "$@")" 
-}
-
 # @cmd Chat with replicate api
 # @env REPLICATE_API_KEY!
 # @option -m --model=meta/meta-llama-3-8b-instruct $REPLICATE_MODEL
@@ -336,22 +324,32 @@ chat-ernie() {
 -d "$(_build_body ernie "$@")"
 }
 
-
 _argc_before() {
     stream="true"
     if [[ -n "$argc_no_stream" ]]; then
         stream="false"
     fi
 }
 
-_openai_models() {
+_retrieve_models() {
     api_base="${api_base:-"$argc_api_base"}"
     api_key="${api_key:-"$argc_api_key"}"
     _wrapper curl "$api_base/models" \
 -H "Authorization: Bearer $api_key" \
 
 }
 
+_retrieve_api_base() {
+    api_base="${platform_config##*,}"
+    if [[ -z "$api_base" ]]; then
+        key="$(echo $argc_platform |  tr '[:lower:]' '[:upper:]')_API_BASE"
+        api_base="${!key}"
+        if [[ -z "$api_base" ]]; then
+            _die "Miss api_base for $argc_platform; please set $key"
+        fi
+    fi
+}
+
 _choice_model() {
     aichat --list-models
 }
@@ -436,7 +434,7 @@ _build_body() {
     "safetySettings":[{"category":"HARM_CATEGORY_HARASSMENT","threshold":"BLOCK_ONLY_HIGH"},{"category":"HARM_CATEGORY_HATE_SPEECH","threshold":"BLOCK_ONLY_HIGH"},{"category":"HARM_CATEGORY_SEXUALLY_EXPLICIT","threshold":"BLOCK_ONLY_HIGH"},{"category":"HARM_CATEGORY_DANGEROUS_CONTENT","threshold":"BLOCK_ONLY_HIGH"}]
 }'
             ;;
-        ernie|cloudflare)
+        ernie)
             echo '{
     "messages": [
         {

config.example.yaml

@@ -223,10 +223,10 @@ clients:
     region: xxx
 
   # See https://developers.cloudflare.com/workers-ai/
-  - type: cloudflare
-    account_id: xxx
+  - type: openai-compatible
+    name: cloudflare
+    api_base: https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/v1
     api_key: xxx
-    api_base: https://api.cloudflare.com/client/v4    # Optional
 
   # See https://replicate.com/docs
   - type: replicate

models.yaml

@@ -520,6 +520,7 @@
 
 # Links:
 #  - https://developers.cloudflare.com/workers-ai/models/
+#  - https://developers.cloudflare.com/workers-ai/configuration/open-ai-compatibility/
 - platform: cloudflare
   models:
     - name: '@cf/meta/llama-3.1-8b-instruct'

src/client/common.rs

@@ -364,29 +364,32 @@ pub fn create_config(prompts: &[PromptAction], client: &str) -> Result<(String,
 
 pub fn create_openai_compatible_client_config(client: &str) -> Result<Option<(String, Value)>> {
     match super::OPENAI_COMPATIBLE_PLATFORMS
-        .iter()
+        .into_iter()
         .find(|(name, _)| client == *name)
     {
         None => Ok(None),
         Some((name, api_base)) => {
             let mut config = json!({
                 "type": OpenAICompatibleClient::NAME,
                 "name": name,
-                "api_base": api_base,
             });
-            let prompts = if ALL_MODELS.iter().any(|v| &v.platform == name) {
-                vec![("api_key", "API Key:", false, PromptKind::String)]
+            let mut prompts = vec![];
+            if api_base.is_empty() {
+                prompts.push(("api_base", "API Base:", true, PromptKind::String));
             } else {
-                vec![
-                    ("api_key", "API Key:", false, PromptKind::String),
+                config["api_base"] = api_base.into();
+            }
+            prompts.push(("api_key", "API Key:", false, PromptKind::String));
+            if !ALL_MODELS.iter().any(|v| v.platform == name) {
+                prompts.extend([
                     ("models[].name", "Model Name:", true, PromptKind::String),
                     (
                         "models[].max_input_tokens",
                         "Max Input Tokens:",
                         false,
                         PromptKind::Integer,
                     ),
-                ]
+                ]);
             };
             let mut model = client.to_string();
             set_client_config(&prompts, &mut model, &mut config)?;

src/client/gemini.rs

@@ -34,7 +34,7 @@ impl_client_trait!(
         gemini_chat_completions,
         gemini_chat_completions_streaming
     ),
-    (prepare_embeddings, gemini_embeddings),
+    (prepare_embeddings, embeddings),
     (noop_prepare_rerank, noop_rerank),
 );
 
@@ -95,7 +95,7 @@ fn prepare_embeddings(self_: &GeminiClient, data: EmbeddingsData) -> Result<Requ
     Ok(request_data)
 }
 
-async fn gemini_embeddings(builder: RequestBuilder, _model: &Model) -> Result<EmbeddingsOutput> {
+async fn embeddings(builder: RequestBuilder, _model: &Model) -> Result<EmbeddingsOutput> {
     let res = builder.send().await?;
     let status = res.status();
     let data: Value = res.json().await?;

src/client/mod.rs

@@ -33,13 +33,13 @@ register_client!(
     ),
     (vertexai, "vertexai", VertexAIConfig, VertexAIClient),
     (bedrock, "bedrock", BedrockConfig, BedrockClient),
-    (cloudflare, "cloudflare", CloudflareConfig, CloudflareClient),
     (replicate, "replicate", ReplicateConfig, ReplicateClient),
     (ernie, "ernie", ErnieConfig, ErnieClient),
 );
 
-pub const OPENAI_COMPATIBLE_PLATFORMS: [(&str, &str); 18] = [
+pub const OPENAI_COMPATIBLE_PLATFORMS: [(&str, &str); 19] = [
     ("ai21", "https://api.ai21.com/studio/v1"),
+    ("cloudflare", ""),
     ("deepinfra", "https://api.deepinfra.com/v1/openai"),
     ("deepseek", "https://api.deepseek.com"),
     ("fireworks", "https://api.fireworks.ai/inference/v1"),