Agentic extraction better retry (#138)

* bedrock utils for retry

* invoke extraction with long retries

* caching the first prompt

* add review agent model config

* refactor extraction service

* fix pipeline ci failure

* deps

* bug fixes

* pydantic validator update

* pass the linting

* default to empty string

* deps fix

* disable ci comments

* fixes

* json mode for model dump

* add additional retryable errors

* additional retryable errors

* linting fix

Author: kazmer97

.github/workflows/developer-tests.yml

@@ -59,7 +59,7 @@ jobs:
 
       - name: Install Node.js and basedpyright
         run: |
-          curl -fsSL https://deb.nodesource.com/setup_20.x | bash -
+          curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
           apt-get install -y nodejs
           npm install -g basedpyright
 
@@ -104,6 +104,7 @@ jobs:
         with:
           files: lib/idp_common_pkg/test-reports/test-results.xml
           check_name: Test Results
+          comment_mode: off  # Disable PR comments to avoid permission issues on fork PRs
 
       - name: Code Coverage Report
         uses: irongut/CodeCoverageSummary@v1.3.0

config_library/pattern-2/lending-package-sample/config.yaml

@@ -1811,7 +1811,7 @@ agents:
     parameters:
       max_log_events: 5
       time_range_hours_default: 24
-  
+
   chat_companion:
     model_id: us.anthropic.claude-haiku-4-5-20251001-v1:0
 pricing:

lib/idp_common_pkg/idp_common/config/models.py

@@ -19,7 +19,15 @@
 """
 
 from typing import Any, Dict, List, Optional, Union, Literal, Annotated
-from pydantic import BaseModel, ConfigDict, Field, field_validator, Discriminator
+from typing_extensions import Self
+from pydantic import (
+    BaseModel,
+    ConfigDict,
+    Field,
+    field_validator,
+    Discriminator,
+    model_validator,
+)
 
 
 class ImageConfig(BaseModel):
@@ -78,6 +86,10 @@ class AgenticConfig(BaseModel):
 
     enabled: bool = Field(default=False, description="Enable agentic extraction")
     review_agent: bool = Field(default=False, description="Enable review agent")
+    review_agent_model: str | None = Field(
+        default=None,
+        description="Model used for reviewing and correcting extraction work",
+    )
 
 
 class ExtractionConfig(BaseModel):
@@ -120,6 +132,14 @@ def parse_int(cls, v: Any) -> int:
             return int(v) if v else 0
         return int(v)
 
+    @model_validator(mode="after")
+    def set_default_review_agent_model(self) -> Self:
+        """Set review_agent_model to extraction model if not specified."""
+        if not self.agentic.review_agent_model:
+            self.agentic.review_agent_model = self.model
+
+        return self
+
 
 class ClassificationConfig(BaseModel):
     """Document classification configuration"""
@@ -423,7 +443,7 @@ class ErrorAnalyzerConfig(BaseModel):
             "AccessDenied",
             "ThrottlingException",
         ],
-        description="Error patterns to search for in logs"
+        description="Error patterns to search for in logs",
     )
     system_prompt: str = Field(
         default="""
@@ -511,11 +531,10 @@ class ErrorAnalyzerConfig(BaseModel):
                       - No time specified: 24 hours (default)
               
               IMPORTANT: Do not include any search quality reflections, search quality scores, or meta-analysis sections in your response. Only provide the three required sections: Root Cause, Recommendations, and Evidence.""",
-        description="System prompt for error analyzer"
+        description="System prompt for error analyzer",
     )
     parameters: ErrorAnalyzerParameters = Field(
-        default_factory=ErrorAnalyzerParameters,
-        description="Error analyzer parameters"
+        default_factory=ErrorAnalyzerParameters, description="Error analyzer parameters"
     )
 
 
@@ -635,12 +654,10 @@ class AgentsConfig(BaseModel):
     """Agents configuration"""
 
     error_analyzer: Optional[ErrorAnalyzerConfig] = Field(
-        default_factory=ErrorAnalyzerConfig,
-        description="Error analyzer configuration"
+        default_factory=ErrorAnalyzerConfig, description="Error analyzer configuration"
     )
     chat_companion: Optional[ChatCompanionConfig] = Field(
-        default_factory=ChatCompanionConfig,
-        description="Chat companion configuration"
+        default_factory=ChatCompanionConfig, description="Chat companion configuration"
     )