feat: add type checker to api backend & align litellm and old backend (#647)

* move cache auto continue and retry to all api backend

* add type checker to json mode output

* fix CI

* feat: Add json_mode handling and streaming support in chat completion function

* lint

* fix a bug when returning a dict which value could contain int or bool

* remove litellm

---------

Co-authored-by: Xu Yang <xuyang1@microsoft.com>
Co-authored-by: Young <afe.young@gmail.com>

rdagent/components/coder/factor_coder/eva_utils.py

@@ -2,7 +2,7 @@ import io
 import json
 from abc import abstractmethod
 from pathlib import Path
-from typing import Tuple
+from typing import Dict, Tuple
 
 import pandas as pd
 from jinja2 import Environment, StrictUndefined
@@ -212,7 +212,10 @@ class FactorOutputFormatEvaluator(FactorEvaluator):
             try:
                 api = APIBackend() if attempts == 0 else APIBackend(use_chat_cache=False)
                 resp = api.build_messages_and_create_chat_completion(
-                    user_prompt=gen_df_info_str, system_prompt=system_prompt, json_mode=True
+                    user_prompt=gen_df_info_str,
+                    system_prompt=system_prompt,
+                    json_mode=True,
+                    json_target_type=Dict[str, str | bool | int],
                 )
                 resp_dict = json.loads(resp)
                 resp_dict["output_format_decision"] = str(resp_dict["output_format_decision"]).lower() in ["true", "1"]
@@ -556,6 +559,7 @@ class FactorFinalDecisionEvaluator(FactorEvaluator):
                         system_prompt=system_prompt,
                         json_mode=True,
                         seed=attempts,  # in case of useless retrying when cache enabled.
+                        json_target_type=Dict[str, str | bool | int],
                     ),
                 )
                 final_decision = final_evaluation_dict["final_decision"]

rdagent/components/coder/factor_coder/evolving_strategy.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import json
 from pathlib import Path
+from typing import Dict
 
 from jinja2 import Environment, StrictUndefined
 
@@ -168,7 +169,10 @@ class FactorMultiProcessEvolvingStrategy(MultiProcessEvolvingStrategy):
                     APIBackend(
                         use_chat_cache=FACTOR_COSTEER_SETTINGS.coder_use_cache
                     ).build_messages_and_create_chat_completion(
-                        user_prompt=user_prompt, system_prompt=system_prompt, json_mode=True
+                        user_prompt=user_prompt,
+                        system_prompt=system_prompt,
+                        json_mode=True,
+                        json_target_type=Dict[str, str],
                     )
                 )["code"]
                 return code