feat: update cohere api to v2 (#1041)

Argcfile.sh

@@ -250,7 +250,7 @@ chat-claude() {
 # @flag -S --no-stream
 # @arg text~
 chat-cohere() {
-    _wrapper curl -i https://api.cohere.ai/v1/chat \
+    _wrapper curl -i https://api.cohere.ai/v2/chat \
 -X POST \
 -H 'Content-Type: application/json' \
 -H "Authorization: Bearer $COHERE_API_KEY" \
@@ -398,7 +398,7 @@ _build_body() {
     else
         shift
         case "$kind" in
-        openai)
+        openai|cohere)
             echo '{
     "model": "'$argc_model'",
     "messages": [
@@ -408,13 +408,6 @@ _build_body() {
         }
     ],
     "stream": '$stream'
-}'
-            ;;
-        cohere)
-            echo '{
-  "model": "'$argc_model'",
-  "message": "'"$*"'",
-  "stream": '$stream'
 }'
             ;;
         claude)

config.example.yaml

@@ -181,7 +181,7 @@ clients:
 
   # See https://docs.cohere.com/docs/the-cohere-platform
   - type: cohere
-    api_base: https://api.cohere.ai/v1                # Optional
+    api_base: https://api.cohere.ai/v2                # Optional
     api_key: xxx
 
   # See https://docs.perplexity.ai/docs/getting-started

models.yaml

@@ -283,12 +283,27 @@
       max_tokens_per_chunk: 512
       default_chunk_size: 1000
       max_batch_size: 96
+    - name: embed-english-light-v3.0
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 512
+      default_chunk_size: 700
+      max_batch_size: 96
     - name: embed-multilingual-v3.0
       type: embedding
       input_price: 0.1
       max_tokens_per_chunk: 512
       default_chunk_size: 1000
       max_batch_size: 96
+    - name: embed-multilingual-light-v3.0
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 512
+      default_chunk_size: 700
+      max_batch_size: 96
+    - name: rerank-v3.5
+      type: reranker
+      max_input_tokens: 4096
     - name: rerank-english-v3.0
       type: reranker
       max_input_tokens: 4096