Merge pull request #1392 from Codium-ai/tr/model_weak

docs: remove model_weak configuration and simplify model selection

docs/docs/usage-guide/changing_a_model.md

@@ -5,7 +5,6 @@ To use a different model than the default (GPT-4), you need to edit in the [conf
 ```
 [config]
 model = "..."
-model_weak = "..."
 fallback_models = ["..."]
 ```
 
@@ -27,9 +26,8 @@ deployment_id = ""  # The deployment name you chose when you deployed the engine
 and set in your configuration file:
 ```
 [config]
-model="" # the OpenAI model you've deployed on Azure (e.g. gpt-3.5-turbo)
-model_weak="" # the OpenAI model you've deployed on Azure (e.g. gpt-3.5-turbo)
-fallback_models=["..."] # the OpenAI model you've deployed on Azure (e.g. gpt-3.5-turbo)
+model="" # the OpenAI model you've deployed on Azure (e.g. gpt-4o)
+fallback_models=["..."]
 ```
 
 ### Hugging Face
@@ -52,7 +50,6 @@ MAX_TOKENS={
 
 [config] # in configuration.toml
 model = "ollama/llama2"
-model_weak = "ollama/llama2"
 fallback_models=["ollama/llama2"]
 
 [ollama] # in .secrets.toml
@@ -76,7 +73,6 @@ MAX_TOKENS={
 }
 [config] # in configuration.toml
 model = "huggingface/meta-llama/Llama-2-7b-chat-hf"
-model_weak = "huggingface/meta-llama/Llama-2-7b-chat-hf"
 fallback_models=["huggingface/meta-llama/Llama-2-7b-chat-hf"]
 
 [huggingface] # in .secrets.toml
@@ -91,7 +87,6 @@ To use Llama2 model with Replicate, for example, set:
 ```
 [config] # in configuration.toml
 model = "replicate/llama-2-70b-chat:2c1608e18606fad2812020dc541930f2d0495ce32eee50074220b87300bc16e1"
-model_weak = "replicate/llama-2-70b-chat:2c1608e18606fad2812020dc541930f2d0495ce32eee50074220b87300bc16e1"
 fallback_models=["replicate/llama-2-70b-chat:2c1608e18606fad2812020dc541930f2d0495ce32eee50074220b87300bc16e1"]
 [replicate] # in .secrets.toml
 key = ...
@@ -107,7 +102,6 @@ To use Llama3 model with Groq, for example, set:
 ```
 [config] # in configuration.toml
 model = "llama3-70b-8192"
-model_weak = "llama3-70b-8192"
 fallback_models = ["groq/llama3-70b-8192"]
 [groq] # in .secrets.toml
 key = ... # your Groq api key
@@ -121,7 +115,6 @@ To use Google's Vertex AI platform and its associated models (chat-bison/codecha
 ```
 [config] # in configuration.toml
 model = "vertex_ai/codechat-bison"
-model_weak = "vertex_ai/codechat-bison"
 fallback_models="vertex_ai/codechat-bison"
 
 [vertexai] # in .secrets.toml
@@ -140,7 +133,6 @@ To use [Google AI Studio](https://aistudio.google.com/) models, set the relevant
 ```toml
 [config] # in configuration.toml
 model="google_ai_studio/gemini-1.5-flash"
-model_weak="google_ai_studio/gemini-1.5-flash"
 fallback_models=["google_ai_studio/gemini-1.5-flash"]
 
 [google_ai_studio] # in .secrets.toml
@@ -156,7 +148,6 @@ To use Anthropic models, set the relevant models in the configuration section of
 ```
 [config]
 model="anthropic/claude-3-opus-20240229"
-model_weak="anthropic/claude-3-opus-20240229"
 fallback_models=["anthropic/claude-3-opus-20240229"]
 ```
 
@@ -173,7 +164,6 @@ To use Amazon Bedrock and its foundational models, add the below configuration:
 ```
 [config] # in configuration.toml
 model="bedrock/anthropic.claude-3-sonnet-20240229-v1:0"
-model_weak="bedrock/anthropic.claude-3-sonnet-20240229-v1:0"
 fallback_models=["bedrock/anthropic.claude-v2:1"]
 ```
 
@@ -195,7 +185,6 @@ If the relevant model doesn't appear [here](https://github.com/Codium-ai/pr-agen
 ```
 [config]
 model="custom_model_name"
-model_weak="custom_model_name"
 fallback_models=["custom_model_name"]
 ```
 (2) Set the maximal tokens for the model:

pr_agent/algo/pr_processing.py

@@ -11,7 +11,7 @@ from pr_agent.algo.git_patch_processing import (
 from pr_agent.algo.language_handler import sort_files_by_main_languages
 from pr_agent.algo.token_handler import TokenHandler
 from pr_agent.algo.types import EDIT_TYPE, FilePatchInfo
-from pr_agent.algo.utils import ModelType, clip_tokens, get_max_tokens
+from pr_agent.algo.utils import ModelType, clip_tokens, get_max_tokens, get_weak_model
 from pr_agent.config_loader import get_settings
 from pr_agent.git_providers.git_provider import GitProvider
 from pr_agent.log import get_logger
@@ -333,7 +333,7 @@ def generate_full_patch(convert_hunks_to_line_numbers, file_dict, max_tokens_mod
     return total_tokens, patches, remaining_files_list_new, files_in_patch_list
 
 
-async def retry_with_fallback_models(f: Callable, model_type: ModelType = ModelType.WEAK):
+async def retry_with_fallback_models(f: Callable, model_type: ModelType = ModelType.REGULAR):
     all_models = _get_all_models(model_type)
     all_deployments = _get_all_deployments(all_models)
     # try each (model, deployment_id) pair until one is successful, otherwise raise exception
@@ -354,8 +354,8 @@ async def retry_with_fallback_models(f: Callable, model_type: ModelType = ModelT
 
 
 def _get_all_models(model_type: ModelType = ModelType.REGULAR) -> List[str]:
-    if get_settings().config.get('model_weak') and model_type == ModelType.WEAK:
-        model = get_settings().config.model_weak
+    if model_type == ModelType.WEAK:
+        model = get_weak_model()
     else:
         model = get_settings().config.model
     fallback_models = get_settings().config.fallback_models

pr_agent/algo/utils.py

@@ -27,6 +27,12 @@ from pr_agent.config_loader import get_settings, global_settings
 from pr_agent.log import get_logger
 
 
+def get_weak_model() -> str:
+    if get_settings().get("config.model_weak"):
+        return get_settings().config.model_weak
+    return get_settings().config.model
+
+
 class Range(BaseModel):
     line_start: int  # should be 0-indexed
     line_end: int

pr_agent/settings/configuration.toml

@@ -1,8 +1,8 @@
 [config]
 # models
-model_weak="gpt-4o-mini-2024-07-18"
 model="gpt-4o-2024-11-20"
 fallback_models=["gpt-4o-2024-08-06"]
+#model_weak="gpt-4o-mini-2024-07-18" # optional, a weaker model to use for some easier tasks
 # CLI
 git_provider="github"
 publish_output=true