Added support to pass functions to index and llm complete functions (#79

)

* Added support to pass functions to index and llm complete functions

Signed-off-by: Deepak <[email protected]>

* Added docstring

Signed-off-by: Deepak <[email protected]>

---------

Signed-off-by: Deepak <[email protected]>
Signed-off-by: Deepak K <[email protected]>
Co-authored-by: Gayathri <[email protected]>

src/unstract/sdk/index.py

@@ -1,5 +1,6 @@
 import json
-from typing import Any, Optional
+import logging
+from typing import Any, Callable, Optional
 
 from llama_index.core import Document
 from llama_index.core.node_parser import SimpleNodeParser
@@ -25,6 +26,8 @@
 from unstract.sdk.vector_db import VectorDB
 from unstract.sdk.x2txt import X2Text
 
+logger = logging.getLogger(__name__)
+
 
 class Constants:
     TOP_K = 5
@@ -101,27 +104,6 @@ def query_index(
         finally:
             vector_db.close()
 
-    def _cleanup_text(self, full_text):
-        # Remove text which is not required
-        full_text_lines = full_text.split("\n")
-        new_context_lines = []
-        empty_line_count = 0
-        for line in full_text_lines:
-            if line.strip() == "":
-                empty_line_count += 1
-            else:
-                if empty_line_count >= 3:
-                    empty_line_count = 3
-                for i in range(empty_line_count):
-                    new_context_lines.append("")
-                empty_line_count = 0
-                new_context_lines.append(line.rstrip())
-        self.tool.stream_log(
-            f"Old context length: {len(full_text_lines)}, "
-            f"New context length: {len(new_context_lines)}"
-        )
-        return "\n".join(new_context_lines)
-
     def index(
         self,
         tool_id: str,
@@ -136,6 +118,7 @@ def index(
         output_file_path: Optional[str] = None,
         enable_highlight: bool = False,
         usage_kwargs: dict[Any, Any] = {},
+        process_text: Optional[Callable[[str], str]] = None,
     ) -> str:
         """Indexes an individual file using the passed arguments.
 
@@ -276,10 +259,17 @@ def index(
             except AdapterError as e:
                 # Wrapping AdapterErrors with SdkError
                 raise IndexingError(str(e)) from e
+            if process_text:
+                try:
+                    result = process_text(extracted_text)
+                    if isinstance(result, str):
+                        extracted_text = result
+                except Exception as e:
+                    logger.error(f"Error occured inside function 'process_text': {e}")
             full_text.append(
                 {
                     "section": "full",
-                    "text_contents": self._cleanup_text(extracted_text),
+                    "text_contents": extracted_text,
                 }
             )
 

src/unstract/sdk/llm.py

@@ -1,6 +1,6 @@
 import logging
 import re
-from typing import Any, Optional
+from typing import Any, Callable, Optional
 
 from llama_index.core.base.llms.types import CompletionResponseGen
 from llama_index.core.llms import LLM as LlamaIndexLLM
@@ -69,15 +69,41 @@ def _initialise(self):
     def complete(
         self,
         prompt: str,
-        retries: int = 3,
+        process_text: Optional[Callable[[str], str]] = None,
         **kwargs: Any,
     ) -> Optional[dict[str, Any]]:
+        """Generates a completion response for the given prompt.
+
+        Args:
+            prompt (str): The input text prompt for generating the completion.
+            process_text (Optional[Callable[[str], str]], optional): A callable that
+                processes the generated text and extracts specific information.
+                Defaults to None.
+            **kwargs (Any): Additional arguments passed to the completion function.
+
+        Returns:
+            Optional[dict[str, Any]]: A dictionary containing the result of the
+                completion and processed output or None if the completion fails.
+
+        Raises:
+            Any: If an error occurs during the completion process, it will be
+                raised after being processed by `parse_llm_err`.
+        """
         try:
             response: CompletionResponse = self._llm_instance.complete(prompt, **kwargs)
+            process_text_output = {}
+            if process_text:
+                try:
+                    process_text_output = process_text(response, LLM.json_regex)
+                    if not isinstance(process_text_output, dict):
+                        process_text_output = {}
+                except Exception as e:
+                    logger.error(f"Error occured inside function 'process_text': {e}")
+                    process_text_output = {}
             match = LLM.json_regex.search(response.text)
             if match:
                 response.text = match.group(0)
-            return {LLM.RESPONSE: response}
+            return {LLM.RESPONSE: response, **process_text_output}
         except Exception as e:
             raise parse_llm_err(e) from e