Agent-Zero 2.12 new chat Question: How are your status ? Error: UnicodeEncodeError: 'utf-8' codec can't encode characters in position 1494-1496: surrogates not allowed Traceback (most recent call last): Traceback (most recent call last): File "/a0/helpers/extension.py", line 181, in _run_async data["result"] = await data["result"] ^^^^^^^^^^^^^^^^^^^^ File "/a0/agent.py", line 649, in handle_exception raise exception # exception handling is done by extensions ^^^^^^^^^^^^^^^ File "/a0/agent.py", line 425, in monologue prompt = await self.prepare_prompt(loop_data=self.loop_data) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/a0/helpers/extension.py", line 188, in _run_async result = _process_result(data) ^^^^^^^^^^^^^^^^^^^^^ File "/a0/helpers/extension.py", line 148, in _process_result raise exc File "/a0/helpers/extension.py", line 181, in _run_async data["result"] = await data["result"] ^^^^^^^^^^^^^^^^^^^^ File "/a0/agent.py", line 579, in prepare_prompt await extension.call_extensions_async( File "/a0/helpers/extension.py", line 238, in call_extensions_async result = cls(agent=agent).execute(**kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/a0/plugins/_context_window/extensions/python/message_loop_prompts_after/_99_capture_context_usage.py", line 8, in execute capture_context(self.agent, loop_data) File "/a0/plugins/_context_window/helpers/usage.py", line 75, in capture_context parts["extras"] = _output_tokens(agent, "extras", protocol + extras) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/a0/plugins/_context_window/helpers/usage.py", line 205, in _output_tokens return _cached_tokens(agent, cache_key, text) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/a0/plugins/_context_window/helpers/usage.py", line 215, in _cached_tokens

## Title: UnicodeEncodeError: 'utf-8' codec can't encode characters (surrogates not allowed) in _cached_tokens, responses_history and httpx ## Description: When Agent-Zero encounters unescaped terminal sequences, broken JSON outputs, or corrupted characters from tool executions, isolated high/low surrogate pairs (range U+D800 to U+DFFF) enter the message history. This causes multiple sequential UnicodeEncodeError: 'utf-8' codec can't encode characters ... surrogates not allowed crashes across the entire framework because raw .encode("utf-8") calls are used without error handling on strings containing these surrogates. ## Steps to Reproduce: 1. Run a tool or terminal command that returns raw binary data, certain emojis, or non-standard ANSI control codes into the agent loop. 2. Let the context manager try to cache or token-count the response. 3. Observe the crash chain. ## Traceback Summary: The crash cascades through three core locations: 1. Plugin Context Window: _context_window/helpers/usage.py inside _cached_tokens via hashlib.sha256(text.encode("utf-8")). 2. History Component: helpers/responses_history.py inside _json_bytes via json.dumps(...).encode(). 3. HTTP Transport Layer: httpx internally crashes in _content.py via encode_json because litellm forwards the uncleaned surrogate strings to the external API call. ## Proposed Fixes / Workarounds: To prevent the framework from crashing on bad tool outputs, errors="surrogateescape" or errors="replace" should be added to the encoding/decoding chains. * Fix 1 (Context Caching) in /a0/plugins/_context_window/helpers/usage.py: digest = hashlib.sha256(text.encode("utf-8", errors="surrogateescape")).hexdigest() * Fix 2 (History Hashing) in /a0/helpers/responses_history.py: return json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(",", ":")).encode("utf-8", errors="surrogateescape")

Docker are the feature not a bug. It make agent-zero save for me. it runs on my server and one on my laptop. I am very habby about agent-zero 😀

powerfull update, i like it 😀