Extracts candidate values from documents (#226)

Using a separate LLM call per document. The extracted, "candidate" field
values are then provided to the guided conversation. The guided
conversation then works through asking the user which values to use.

Additionally, removes unneeded truncation of completion messages in
message debug, now that it's stored and retrieved separately.

assistants/explorer-assistant/assistant/chat.py

@@ -334,7 +334,7 @@ async def respond_to_conversation(
                         method_metadata_key: {
                             "request": {
                                 "model": config.request_config.openai_model,
-                                "messages": openai_client.truncate_messages_for_logging(completion_messages),
+                                "messages": completion_messages,
                                 "max_tokens": config.request_config.response_tokens,
                             },
                             "response": completion.model_dump() if completion else "[no response from openai]",
@@ -358,7 +358,7 @@ async def respond_to_conversation(
                         method_metadata_key: {
                             "request": {
                                 "model": config.request_config.openai_model,
-                                "messages": openai_client.truncate_messages_for_logging(completion_messages),
+                                "messages": completion_messages,
                             },
                             "error": str(e),
                         },

assistants/prospector-assistant/assistant/agents/document_agent.py

@@ -919,7 +919,7 @@ def _on_success_metadata_update(
                 f"{method_metadata_key}": {
                     "request": {
                         "model": config.request_config.openai_model,
-                        "messages": openai_client.truncate_messages_for_logging(chat_completion_messages),
+                        "messages": chat_completion_messages,
                         "max_tokens": config.request_config.response_tokens,
                     },
                     "response": completion.model_dump() if completion else "[no response from openai]",
@@ -943,7 +943,7 @@ def _on_error_metadata_update(
                 f"{method_metadata_key}": {
                     "request": {
                         "model": config.request_config.openai_model,
-                        "messages": openai_client.truncate_messages_for_logging(chat_completion_messages),
+                        "messages": chat_completion_messages,
                     },
                     "error": str(e),
                 },