Move provider definitions to separate package

gptcli/assistant.py

@@ -3,17 +3,17 @@
 from attr import dataclass
 import platform
 from typing import Any, Dict, Iterator, Optional, TypedDict, List
-from gptcli.cohere import CohereCompletionProvider
 
 from gptcli.completion import (
     CompletionEvent,
     CompletionProvider,
     ModelOverrides,
     Message,
 )
-from gptcli.llama import LLaMACompletionProvider
-from gptcli.openai import OpenAICompletionProvider
-from gptcli.anthropic import AnthropicCompletionProvider
+from gptcli.providers.llama import LLaMACompletionProvider
+from gptcli.providers.openai import OpenAICompletionProvider
+from gptcli.providers.anthropic import AnthropicCompletionProvider
+from gptcli.providers.cohere import CohereCompletionProvider
 
 
 class AssistantConfig(TypedDict, total=False):

gptcli/config.py

@@ -4,7 +4,7 @@
 import yaml
 
 from gptcli.assistant import AssistantConfig
-from gptcli.llama import LLaMAModelConfig
+from gptcli.providers.llama import LLaMAModelConfig
 
 
 CONFIG_FILE_PATHS = [

gptcli/gpt.py

@@ -13,8 +13,8 @@
 import sys
 import logging
 import datetime
-import gptcli.anthropic
-import gptcli.cohere
+import gptcli.providers.anthropic
+import gptcli.providers.cohere
 from gptcli.assistant import (
     Assistant,
     DEFAULT_ASSISTANTS,
@@ -32,7 +32,7 @@
     choose_config_file,
     read_yaml_config,
 )
-from gptcli.llama import init_llama_models
+from gptcli.providers.llama import init_llama_models
 from gptcli.logging import LoggingChatListener
 from gptcli.cost import PriceChatListener
 from gptcli.session import ChatSession
@@ -184,10 +184,10 @@ def main():
         openai.api_key = config.openai_api_key
 
     if config.anthropic_api_key:
-        gptcli.anthropic.api_key = config.anthropic_api_key
+        gptcli.providers.anthropic.api_key = config.anthropic_api_key
 
     if config.cohere_api_key:
-        gptcli.cohere.api_key = config.cohere_api_key
+        gptcli.providers.cohere.api_key = config.cohere_api_key
 
     if config.llama_models is not None:
         init_llama_models(config.llama_models)

gptcli/shell.py

@@ -14,8 +14,9 @@ def simple_response(assistant: Assistant, prompt: str, stream: bool) -> None:
     result = ""
     try:
         for response in response_iter:
-            result += response
-            sys.stdout.write(response)
+            if response.type == "message_delta":
+                result += response.text
+                sys.stdout.write(response.text)
     except KeyboardInterrupt:
         pass
     finally:
@@ -29,6 +30,8 @@ def execute(assistant: Assistant, prompt: str) -> None:
     logging.info("User: %s", prompt)
     response_iter = assistant.complete_chat(messages, stream=False)
     result = next(response_iter)
+    assert result.type == "message_delta"
+    result = result.text
     logging.info("Assistant: %s", result)
 
     with tempfile.NamedTemporaryFile(mode="w", prefix="gptcli-", delete=False) as f: