Integrate agent trigger with Sandbox HTTP API

installer/slackbot/AGENT_COMMAND_DOCS.md

@@ -0,0 +1,53 @@
+# Slack Bot Agent Command Documentation
+
+## Overview
+
+The `slack_bot.py` script runs a Socket Mode client that listens for commands in a specific Slack channel. The core integration point for external agent interaction is the `!agent` command, managed by the `handle_agent` function.
+
+## The `!agent` Command Flow
+
+When a user types `!agent <instructions>` (e.g., `!agent analyze the logs`), the bot performs the following steps:
+
+1.  **Input Parsing:** 
+    - It strips the `!agent` prefix to isolate the `instructions` string.
+    - If no instructions are provided, it returns a warning message.
+
+2.  **Local Persistence:** 
+    - The instructions are written to a local file defined by `AGENT_PAYLOAD_PATH` (default: `agent_payload.txt`).
+
+3.  **HTTP POST Request:** 
+    - The bot sends a POST request to the `SANDBOX_URL` (default: `http://sandbox:8080`).
+    - It sets a timeout of 30 seconds.
+
+4.  **Response Handling:** 
+    - If the request is successful, the bot captures the response.
+    - It attempts to parse the response as JSON; if that fails, it falls back to raw text.
+    - The first 2000 characters of the result are sent back to the Slack channel.
+
+## HTTP Request Structure
+
+### Request Payload
+
+The bot sends the user's instructions as a JSON object with a single key: `prompt`.
+
+**Example Request:**
+
+```json
+{
+  "prompt": "Please analyze the latest server logs and summarize errors."
+}
+```
+
+### Response Expectation
+
+While the bot handles raw text responses, a JSON response is preferred for structured data handling.
+
+**Example Response:**
+
+```json
+{
+  "status": "success",
+  "job_id": "12345",
+  "message": "Analysis started. ETA 2 minutes."
+}
+```

installer/slackbot/slack_bot.py

@@ -22,13 +22,7 @@
 WEBHOOK_URL = os.environ.get("SLACK_WEBHOOK_URL")
 DEFAULT_CHANNEL = os.environ.get("SLACK_DEFAULT_CHANNEL", "C08NTG6CXL5")
 AGENT_PAYLOAD_PATH = Path(os.environ.get("AGENT_PAYLOAD_PATH", "agent_payload.txt"))
-AGENT_TRIGGER_COMMAND = os.environ.get("AGENT_TRIGGER_COMMAND")
-DEFAULT_AGENT_COMMAND = [
-    "python3",
-    "-c",
-    "print('Agent trigger placeholder executed')",
-]
-AGENT_COMMAND = shlex.split(AGENT_TRIGGER_COMMAND) if AGENT_TRIGGER_COMMAND else DEFAULT_AGENT_COMMAND
+SANDBOX_URL = os.environ.get("SANDBOX_URL", "http://sandbox:8080")
 
 
 # --- Public helper functions ---
@@ -110,32 +104,33 @@ def handle_agent(user, command_full):
             text=f"❌ <@{user}> Unable to write agent payload. Error: {exc}",
         )
         return
-    #TODO: Add AI + Terrarium integration here
 
+    # Send payload to Sandbox via HTTP POST
     try:
-        result = subprocess.run(
-            AGENT_COMMAND,
-            capture_output=True,
-            text=True,
-            check=True,
-        )
-        output = (result.stdout or "Command executed with no output").strip()
+        payload = {"prompt": instructions}
+        print(f"Sending agent instructions to {SANDBOX_URL}...")
+        response = requests.post(SANDBOX_URL, json=payload, timeout=30)
+        response.raise_for_status()
+
+        # Try to parse response as JSON, otherwise use text
+        try:
+            resp_data = response.json()
+            output_preview = str(resp_data)[:2000]
+        except ValueError:
+            output_preview = response.text[:2000]
+
         send_slack_message(
             DEFAULT_CHANNEL,
             text=(
-                f"✅ <@{user}> Agent instructions saved to `{AGENT_PAYLOAD_PATH}`."
-                " Placeholder trigger output:\n```${output[:2000]}```"
+                f"✅ <@{user}> Agent instructions sent to Sandbox.\n"
+                f"Response:\n```\n{output_preview}\n```"
             ),
         )
-    except subprocess.CalledProcessError as exc:
-        error_text = (exc.stderr or str(exc)).strip()
-        print("Agent trigger command failed:", error_text)
+    except Exception as exc:
+        print(f"Error communicating with sandbox: {exc}")
         send_slack_message(
             DEFAULT_CHANNEL,
-            text=(
-                f"❌ <@{user}> Agent trigger command failed with exit code {exc.returncode}."
-                f" Details:\n```${error_text[:2000]}```"
-            ),
+            text=f"❌ <@{user}> Failed to trigger Sandbox pipeline. Error: {exc}",
         )