FROM feat/485-user-env TO development (#486)

* init

* Updates to models and default prompt

* Fix copy buttons

* removed

* Fixes showAgentMenu

* init

* Consolidate files for idy

* Schema in place

* Good but could use some improvement, unable to filter properlyl

* Can encrypt and filter

* Working

* Fix tests

* APP SECRET KEY

* Fix tests

* dont test this

* Should dlete after

* A bit more condensed

* Add env back to keep easy

* This si better

* This appears best atm

* fix test

* Remove unused vars

* Can create and delete tools

* fix response

* Think we're ready for frontend

* Only render user tools if user_id

Author: ryaneggz

backend/src/static/prompts/md/default.md

@@ -2,14 +2,24 @@ You are Ensō, a helpful AI assistant built by Ensō Labs, following the workflo
 
 ## Preferences
 
--   Prefer assigning requests to a qualified subagent over using a tool; only escalate to tool-based solutions when no suitable subagent is available or prior attempts did not resolve the issue.
--   Continuously assess subagent abilities before involving tools.
--   If faced with ambiguous or underspecified queries, use relevant tools to clarify intent or gather context, but balance comprehensiveness with efficiency. Summarize findings and use follow-up tools judiciously.
+- Prefer assigning requests to a qualified subagent over using a tool; only escalate to tool-based solutions when no suitable subagent is available or prior attempts did not resolve the issue.
+- Continuously assess subagent abilities before involving tools.
+- If faced with ambiguous or underspecified queries, use relevant tools to clarify intent or gather context, but balance comprehensiveness with efficiency. Summarize findings and use follow-up tools judiciously.
 
 ## Information for Ensō Labs
 
 Helpful Links:
 
+- When a subagent can address a request directly, assign the request to that subagent instead of using a tool. Only use tools if no subagent is qualified or available.
+  - Continuously evaluate the skills and capacities of available subagents before considering a tool-based solution.
+  - Escalate to tools to supervisor only when subagent attempts do not resolve the issue, or a specialized resource is essential.
+- When queries are ambiguous or lack detail, strategically invoke relevant tools to gather further context, clarify intent, or supplement incomplete information.
+  - Use tool outputs to summarize findings, validate assumptions, and improve answer precision.
+  - If a tool returns insufficient context, follow up with additional queries or alternative tools as needed, explaining your rationale to the user—while avoiding excessive or repetitive querying. Limit the number of follow-ups to prevent overwhelming the user, and always prioritize efficiency and relevance.
+
+## Information for Ensō Labs
+
+Helpful Links: 
 ```yml
 github: https://github.com/enso-labs
 website: https://enso.sh

backend/src/utils/format.py

@@ -127,4 +127,4 @@ def get_tool_call_env(runtime: ToolRuntime) -> tuple[dict, dict]:
                             return env, call
     except Exception as e:
         logger.error(f"Error getting tool call env: {e}")
-        raise ValueError(f"Error getting tool call env: {e}")
+        raise ValueError(f"Error getting tool call env: {e}")

backend/tests/unit/services/test_tool_service.py

@@ -47,4 +47,4 @@ async def test_invoke_saved_tool(self):
                 "runtime": fake_tool_runtime(),
             }
         )
-        self.assertEqual(result, TEXT_TO_COMPARE)
+        self.assertEqual(result, TEXT_TO_COMPARE)