@ducci/jarvis 1.0.11 → 1.0.13

package/docs/system-prompt.md

@@ -2,13 +2,17 @@
 
 This is the authoritative system prompt sent to the model at the start of every session. It is stored as the first message (`role: "system"`) in the conversation history.
 
-Before sending to the model, the server replaces the `{{user_info}}` placeholder with the current contents of `user-info.json`. This happens at runtime on every request — the placeholder is never stored in the conversation history.
+Before sending to the model, the server replaces the `{{user_info}}` and `{{session_id}}` placeholders at runtime on every request — these are never stored in the conversation history.
 
 ---
 
 ```
 You are Jarvis, a fully autonomous agent running on a local server. You have access to tools and can execute shell commands on the machine you run on.
 
+## Session
+
+Current session ID: {{session_id}}
+
 ## Known User Context
 
 {{user_info}}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ducci/jarvis",
-  "version": "1.0.11",
+  "version": "1.0.13",
   "description": "A fully automated agent system that lives on a server.",
   "main": "./src/index.js",
   "type": "module",

package/src/server/agent.js

@@ -7,6 +7,7 @@ import { appendLog } from './logging.js';
 import chalk from 'chalk';
 
 const FORMAT_NUDGE = 'Your previous response was not valid JSON. Respond only with the required JSON object: {"response": "...", "logSummary": "..."}';
+const LOOP_DETECTION_THRESHOLD = 3;
 
 const WRAP_UP_NOTE = `[System: You have reached the iteration limit. This is your final response for this run.
 Respond with your normal JSON, but add a checkpoint field:
@@ -69,6 +70,7 @@ async function runAgentLoop(client, config, session, prepareMessages) {
   let toolDefs = getToolDefinitions(tools);
   let iteration = 0;
   const runToolCalls = [];
+  const loopTracker = new Map();
   let done = false;
   let response = '';
   let logSummary = '';
@@ -154,6 +156,17 @@ async function runAgentLoop(client, config, session, prepareMessages) {
           tool_call_id: toolCall.id,
           content: resultStr,
         });
+
+        const callKey = `${toolName}|${JSON.stringify(toolArgs)}|${resultStr}`;
+        loopTracker.set(callKey, (loopTracker.get(callKey) || 0) + 1);
+      }
+
+      const loopDetected = [...loopTracker.values()].some(count => count >= LOOP_DETECTION_THRESHOLD);
+      if (loopDetected) {
+        session.messages.push({
+          role: 'user',
+          content: '[System: Loop detected. You are repeatedly calling the same tools with identical arguments and getting identical results. Stop calling tools and provide your final answer now based on what you already know.]',
+        });
       }
 
       // Reload tools if any were created/updated this iteration
@@ -245,31 +258,47 @@ async function runAgentLoop(client, config, session, prepareMessages) {
       };
     }
 
-    const wrapUpContent = wrapUpResult.choices[0].message.content || '';
-    // Store the wrap-up response (but NOT the temporary system note)
-    session.messages.push({ role: 'assistant', content: wrapUpContent });
+    let wrapUpContent = wrapUpResult.choices[0].message.content || '';
+    let parsedWrapUp = null;
 
+    // Try JSON parse; if it fails, nudge retry (Layer 2)
     try {
-      const parsed = JSON.parse(wrapUpContent);
-      response = parsed.response || '';
-      logSummary = parsed.logSummary || '';
+      parsedWrapUp = JSON.parse(wrapUpContent);
+    } catch {
+      try {
+        const nudgeMessages = [...wrapUpMessages, { role: 'user', content: FORMAT_NUDGE }];
+        const nudgeResult = await callModelWithFallback(client, config, nudgeMessages, []);
+        const nudgeContent = nudgeResult.choices[0]?.message?.content || '';
+        parsedWrapUp = JSON.parse(nudgeContent);
+        wrapUpContent = nudgeContent;
+      } catch {
+        // Layer 3: use raw text as best-effort response below
+      }
+    }
+
+    // Store the wrap-up response (but NOT the temporary system note)
+    session.messages.push({ role: 'assistant', content: wrapUpContent });
 
-      if (parsed.checkpoint) {
+    if (parsedWrapUp) {
+      response = parsedWrapUp.response || '';
+      logSummary = parsedWrapUp.logSummary || '';
+      if (parsedWrapUp.checkpoint) {
         return {
           iteration,
           response,
           logSummary,
           status: 'checkpoint_reached',
           runToolCalls,
-          checkpoint: parsed.checkpoint,
+          checkpoint: parsedWrapUp.checkpoint,
         };
       }
-    } catch {
+      status = 'ok';
+    } else {
+      // Layer 3: use raw text — user gets a real response instead of an error
       response = wrapUpContent;
-      logSummary = 'Wrap-up response was not valid JSON.';
+      logSummary = 'Wrap-up response was not valid JSON after retry.';
+      status = 'ok';
     }
-
-    status = 'checkpoint_reached';
   }
 
   return { iteration, response, logSummary, status, runToolCalls, checkpoint: null };
@@ -301,7 +330,7 @@ export async function handleChat(config, requestSessionId, userMessage) {
   function prepareMessages(messages) {
     return messages.map((msg, i) => {
       if (i === 0 && msg.role === 'system') {
-        return { ...msg, content: resolveSystemPrompt(msg.content) };
+        return { ...msg, content: resolveSystemPrompt(msg.content, sessionId) };
       }
       return msg;
     });

package/src/server/config.js

@@ -63,7 +63,7 @@ export function loadSystemPrompt() {
   return match[1].trim();
 }
 
-export function resolveSystemPrompt(promptTemplate) {
+export function resolveSystemPrompt(promptTemplate, sessionId) {
   let userInfo = '(none yet)';
   try {
     const raw = fs.readFileSync(PATHS.userInfoFile, 'utf8');
@@ -74,5 +74,7 @@ export function resolveSystemPrompt(promptTemplate) {
   } catch {
     // File doesn't exist yet
   }
-  return promptTemplate.replace('{{user_info}}', userInfo);
+  return promptTemplate
+    .replace('{{session_id}}', sessionId || 'unknown')
+    .replace('{{user_info}}', userInfo);
 }