satif-ai 0.2.3__tar.gz → 0.2.5__tar.gz

{satif_ai-0.2.3 → satif_ai-0.2.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: satif-ai
-Version: 0.2.3
+Version: 0.2.5
 Summary: AI Agents for Satif
 License: MIT
 Author: Bryan Djafer

{satif_ai-0.2.3 → satif_ai-0.2.5}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "satif-ai"
-version = "0.2.3"
+version = "0.2.5"
 description = "AI Agents for Satif"
 authors = [
     {name = "Bryan Djafer", email = "bryan.djafer@syncpulse.fr"}
@@ -20,7 +20,7 @@ requires = ["poetry-core>=2.0.0,<3.0.0"]
 build-backend = "poetry.core.masonry.api"
 
 [project.scripts]
-satif = "satif.cli:main"
+satif-ai = "satif.cli:main"
 
 [tool.poetry.group.dev.dependencies]
 pytest = "^8.3.5"

satif_ai-0.2.5/satif_ai/code_builders/transformation.py

@@ -0,0 +1,238 @@
+import base64
+import os
+import re
+from pathlib import Path
+from typing import Dict, List, Optional, Union
+
+from agents import Agent, Runner, function_tool
+from agents.mcp.server import MCPServer
+from mcp import ClientSession
+from satif_core import AsyncCodeBuilder, CodeBuilder, SDIFDatabase
+from satif_sdk.comparators import get_comparator
+from satif_sdk.representers import get_representer
+from satif_sdk.transformers import CodeTransformer
+
+# Global variables for transformation
+INPUT_SDIF_PATH: Optional[Path] = None
+OUTPUT_TARGET_FILES: Optional[Dict[Union[str, Path], str]] = None
+SCHEMA_ONLY: Optional[bool] = None
+
+
+@function_tool
+async def execute_transformation(code: str) -> str:
+    """Executes the transformation code on the input and returns the
+    comparison difference between the transformed output and the target output example.
+
+    Args:
+        code: The code to execute on the input.
+    """
+    if INPUT_SDIF_PATH is None or OUTPUT_TARGET_FILES is None:
+        return "Error: Transformation context not initialized"
+
+    code_transformer = CodeTransformer(function=code)
+    generated_output_path = code_transformer.export(INPUT_SDIF_PATH)
+
+    comparisons = []
+    comparator_kwargs = {}
+    if SCHEMA_ONLY:
+        comparator_kwargs["check_structure_only"] = True
+
+    if os.path.isdir(generated_output_path):
+        # If it's a directory, compare each file with its corresponding target
+        generated_files = os.listdir(generated_output_path)
+
+        for (
+            output_base_file,
+            output_target_file_name,
+        ) in OUTPUT_TARGET_FILES.items():
+            if output_target_file_name in generated_files:
+                generated_file_path = os.path.join(
+                    generated_output_path, output_target_file_name
+                )
+                comparator = get_comparator(output_target_file_name.split(".")[-1])
+                comparison = comparator.compare(
+                    generated_file_path, output_base_file, **comparator_kwargs
+                )
+                comparisons.append(
+                    f"Comparison for {generated_file_path} [SOURCE] with {output_target_file_name} [TARGET]: {comparison}"
+                )
+            else:
+                comparisons.append(
+                    f"Error: {output_target_file_name} not found in the generated output"
+                )
+    else:
+        # If it's a single file, ensure there's only one target and compare
+        if len(OUTPUT_TARGET_FILES) == 1:
+            output_file = list(OUTPUT_TARGET_FILES.keys())[0]
+            output_target_file_name = list(OUTPUT_TARGET_FILES.values())[0]
+            comparator = get_comparator(output_file.split(".")[-1])
+            comparison = comparator.compare(
+                generated_output_path, output_file, **comparator_kwargs
+            )
+            comparisons.append(
+                f"Comparison for {generated_output_path} [SOURCE] with {output_target_file_name} [TARGET]: {comparison}"
+            )
+        else:
+            comparisons.append(
+                "Error: Single output file generated but multiple target files expected"
+            )
+
+    return "\n".join(comparisons)
+
+
+class TransformationCodeBuilder(CodeBuilder):
+    def __init__(self, output_example: Path | List[Path] | Dict[str, Path]):
+        self.output_example = output_example
+
+    def build(
+        self,
+        sdif: Path | SDIFDatabase,
+        instructions: Optional[str] = None,
+    ) -> str:
+        pass
+
+
+class TransformationAsyncCodeBuilder(AsyncCodeBuilder):
+    """This class is used to build a transformation code that will be used to transform a SDIF database into a set of files following the format of the given output files."""
+
+    def __init__(
+        self,
+        mcp_server: MCPServer,
+        mcp_session: ClientSession,
+        llm_model: str = "o4-mini",
+    ):
+        self.mcp_server = mcp_server
+        self.mcp_session = mcp_session
+        self.llm_model = llm_model
+
+    async def build(
+        self,
+        sdif: Path,  # This will now be relative to project root (MCP server CWD)
+        output_target_files: Dict[Union[str, Path], str] | List[Path],
+        output_sdif: Optional[Path] = None,  # This will now be relative or None
+        instructions: Optional[str] = None,
+        schema_only: bool = False,
+    ) -> str:
+        global INPUT_SDIF_PATH, OUTPUT_TARGET_FILES, SCHEMA_ONLY
+        # INPUT_SDIF_PATH is used by execute_transformation tool, needs to be accessible from where that tool runs.
+        # If execute_transformation runs in the same process as the builder, absolute path is fine.
+        # If it were a separate context, this might need adjustment.
+        # For now, assume execute_transformation can access absolute paths if needed for its *input SDIF*.
+        # However, the sdif for MCP URIs must be relative.
+        INPUT_SDIF_PATH = Path(sdif).resolve()
+        SCHEMA_ONLY = schema_only
+        # Paths for MCP URIs are now expected to be relative to MCP server CWD (project root)
+        # So, use them directly as strings.
+        input_sdif_mcp_uri_path = base64.b64encode(str(sdif).encode()).decode()
+        output_sdif_mcp_uri_path = (
+            base64.b64encode(str(output_sdif).encode()).decode()
+            if output_sdif
+            else None
+        )
+
+        input_schema = await self.mcp_session.read_resource(
+            f"schema://{input_sdif_mcp_uri_path}"
+        )
+        input_sample = await self.mcp_session.read_resource(
+            f"sample://{input_sdif_mcp_uri_path}"
+        )
+
+        output_schema_text = "N/A"
+        output_sample_text = "N/A"
+        if output_sdif_mcp_uri_path:
+            try:
+                output_schema_content = await self.mcp_session.read_resource(
+                    f"schema://{output_sdif_mcp_uri_path}"
+                )
+                if output_schema_content.contents:
+                    output_schema_text = output_schema_content.contents[0].text
+            except Exception as e:
+                print(
+                    f"Warning: Could not read schema for output_sdif {output_sdif_mcp_uri_path}: {e}"
+                )
+
+            try:
+                output_sample_content = await self.mcp_session.read_resource(
+                    f"sample://{output_sdif_mcp_uri_path}"
+                )
+                if output_sample_content.contents:
+                    output_sample_text = output_sample_content.contents[0].text
+            except Exception as e:
+                print(
+                    f"Warning: Could not read sample for output_sdif {output_sdif_mcp_uri_path}: {e}"
+                )
+
+        # OUTPUT_TARGET_FILES keys are absolute paths to original example files for local reading by representers/comparators.
+        # Values are agent-facing filenames.
+        if isinstance(output_target_files, list):
+            OUTPUT_TARGET_FILES = {
+                file_path.resolve(): file_path.name for file_path in output_target_files
+            }
+        elif isinstance(output_target_files, dict):
+            temp_map = {}
+            for k, v in output_target_files.items():
+                if isinstance(k, Path):
+                    temp_map[k.resolve()] = v
+                else:
+                    temp_map[k] = v
+            OUTPUT_TARGET_FILES = temp_map
+        else:
+            OUTPUT_TARGET_FILES = {}
+
+        output_representation = {}
+        if OUTPUT_TARGET_FILES:
+            for file_key_abs_path in list(OUTPUT_TARGET_FILES.keys()):
+                agent_facing_name = OUTPUT_TARGET_FILES[file_key_abs_path]
+                print(f"Representing {agent_facing_name} from {file_key_abs_path}!!")
+                try:
+                    # Representer uses the absolute path (file_key_abs_path) to read the example file.
+                    output_representation[agent_facing_name] = get_representer(
+                        file_key_abs_path
+                    ).represent(file_key_abs_path)
+                except Exception as e:
+                    print(
+                        f"Warning: Could not get representation for {agent_facing_name} (path {file_key_abs_path}): {e}"
+                    )
+                    output_representation[agent_facing_name] = (
+                        f"Error representing file: {e}"
+                    )
+
+        prompt = await self.mcp_session.get_prompt(
+            "create_transformation",
+            arguments={
+                "input_file": Path(
+                    input_sdif_mcp_uri_path
+                ).name,  # Display name for prompt (from relative path)
+                "input_schema": input_schema.contents[0].text
+                if input_schema.contents
+                else "Error reading input schema",
+                "input_sample": input_sample.contents[0].text
+                if input_sample.contents
+                else "Error reading input sample",
+                "output_files": str(list(OUTPUT_TARGET_FILES.values())),
+                "output_schema": output_schema_text,
+                "output_sample": output_sample_text
+                if not SCHEMA_ONLY
+                else "Sample not available. Transform according to the schema only. No data is present in the output example.",
+                "output_representation": str(
+                    output_representation
+                ),  # Representation keyed by agent-facing name
+            },
+        )
+        agent = Agent(
+            name="Transformation Builder",
+            mcp_servers=[self.mcp_server],
+            tools=[execute_transformation],
+            model=self.llm_model,
+        )
+        result = await Runner.run(agent, prompt.messages[0].content.text)
+        transformation_code = self.parse_code(result.final_output)
+        return transformation_code
+
+    def parse_code(self, code) -> str:
+        match = re.search(r"```(?:python)?(.*?)```", code, re.DOTALL)
+        if match:
+            return match.group(1).strip()
+        else:
+            # Handle case where no code block is found
+            return code.strip()

satif_ai-0.2.3/satif_ai/code_builders/transformation.py

@@ -1,152 +0,0 @@
-import os
-import re
-from pathlib import Path
-from typing import Dict, List, Optional, Union
-
-from agents import Agent, Runner, function_tool
-from agents.mcp.server import MCPServerStdio
-from mcp import ClientSession
-from satif_core import AsyncCodeBuilder, CodeBuilder, SDIFDatabase
-from satif_sdk.comparators import get_comparator
-from satif_sdk.representers import get_representer
-from satif_sdk.transformers import CodeTransformer
-
-# Global variables for transformation
-INPUT_SDIF_PATH: Optional[Path] = None
-OUTPUT_TARGET_FILES: Optional[Dict[Union[str, Path], str]] = None
-
-
-@function_tool
-async def execute_transformation(code: str) -> str:
-    """Executes the transformation code on the input and returns the
-    comparison difference between the transformed output and the target output example.
-
-    Args:
-        code: The code to execute on the input.
-    """
-    if INPUT_SDIF_PATH is None or OUTPUT_TARGET_FILES is None:
-        return "Error: Transformation context not initialized"
-
-    code_transformer = CodeTransformer(function=code)
-    generated_output_path = code_transformer.export(INPUT_SDIF_PATH)
-
-    comparisons = []
-
-    if os.path.isdir(generated_output_path):
-        # If it's a directory, compare each file with its corresponding target
-        generated_files = os.listdir(generated_output_path)
-
-        for (
-            output_base_file,
-            output_target_file_name,
-        ) in OUTPUT_TARGET_FILES.items():
-            if output_target_file_name in generated_files:
-                generated_file_path = os.path.join(
-                    generated_output_path, output_target_file_name
-                )
-                comparator = get_comparator(output_target_file_name.split(".")[-1])
-                comparison = comparator.compare(generated_file_path, output_base_file)
-                comparisons.append(
-                    f"Comparison for {generated_file_path} [SOURCE] with {output_target_file_name} [TARGET]: {comparison}"
-                )
-            else:
-                comparisons.append(
-                    f"Error: {output_target_file_name} not found in the generated output"
-                )
-    else:
-        # If it's a single file, ensure there's only one target and compare
-        if len(OUTPUT_TARGET_FILES) == 1:
-            output_file = list(OUTPUT_TARGET_FILES.keys())[0]
-            output_target_file_name = list(OUTPUT_TARGET_FILES.values())[0]
-            comparator = get_comparator(output_file.split(".")[-1])
-            comparison = comparator.compare(generated_output_path, output_file)
-            comparisons.append(
-                f"Comparison for {generated_output_path} [SOURCE] with {output_target_file_name} [TARGET]: {comparison}"
-            )
-        else:
-            comparisons.append(
-                "Error: Single output file generated but multiple target files expected"
-            )
-
-    return "\n".join(comparisons)
-
-
-class TransformationCodeBuilder(CodeBuilder):
-    def __init__(self, output_example: Path | List[Path] | Dict[str, Path]):
-        self.output_example = output_example
-
-    def build(
-        self,
-        sdif: Path | SDIFDatabase,
-        instructions: Optional[str] = None,
-    ) -> str:
-        pass
-
-
-class TransformationAsyncCodeBuilder(AsyncCodeBuilder):
-    """This class is used to build a transformation code that will be used to transform a SDIF database into a set of files following the format of the given output files."""
-
-    def __init__(
-        self,
-        mcp_server: MCPServerStdio,
-        mcp_session: ClientSession,
-        llm_model: str = "o3-mini",
-    ):
-        self.mcp_server = mcp_server
-        self.mcp_session = mcp_session
-        self.llm_model = llm_model
-
-    async def build(
-        self,
-        sdif: Path,
-        output_target_files: Dict[Union[str, Path], str] | List[Path],
-        output_sdif: Optional[Path] = None,
-        instructions: Optional[str] = None,
-    ) -> str:
-        global INPUT_SDIF_PATH, OUTPUT_TARGET_FILES
-        INPUT_SDIF_PATH = Path(sdif)
-
-        if isinstance(output_target_files, list):
-            OUTPUT_TARGET_FILES = {file: file.name for file in output_target_files}
-        else:
-            OUTPUT_TARGET_FILES = output_target_files
-
-        input_schema = await self.mcp_session.read_resource(f"schema://{sdif}")
-        input_sample = await self.mcp_session.read_resource(f"sample://{sdif}")
-
-        output_schema = await self.mcp_session.read_resource(f"schema://{output_sdif}")
-        output_sample = await self.mcp_session.read_resource(f"sample://{output_sdif}")
-        output_representation = {
-            file: get_representer(file).represent(file)
-            for file in list(OUTPUT_TARGET_FILES.keys())
-        }
-
-        prompt = await self.mcp_session.get_prompt(
-            "create_transformation",
-            arguments={
-                "input_file": INPUT_SDIF_PATH.name,
-                "input_schema": input_schema.contents[0].text,
-                "input_sample": input_sample.contents[0].text,
-                "output_files": str(list(OUTPUT_TARGET_FILES.values())),
-                "output_schema": output_schema.contents[0].text,
-                "output_sample": output_sample.contents[0].text,
-                "output_representation": str(output_representation),
-            },
-        )
-        agent = Agent(
-            name="Transformation Builder",
-            mcp_servers=[self.mcp_server],
-            tools=[execute_transformation],
-            model=self.llm_model,
-        )
-        result = await Runner.run(agent, prompt.messages[0].content.text)
-        transformation_code = self.parse_code(result.final_output)
-        return transformation_code
-
-    def parse_code(self, code) -> str:
-        match = re.search(r"```(?:python)?(.*?)```", code, re.DOTALL)
-        if match:
-            return match.group(1).strip()
-        else:
-            # Handle case where no code block is found
-            return code.strip()