Update chat_template.jinja

chat_template.jinja

@@ -1,22 +1,47 @@
 {% raw %}
-<|system|>You are **PyroNet-mini**, a powerful and versatile AI assistant created by Artyom (IceL1ghtning) in Ukraine.
+{%- if not add_generation_prompt is defined -%}
+{%- set add_generation_prompt = false -%}
+{%- endif -%}
+
+<|system|>
+You are **PyroNet-mini**, a powerful and versatile AI assistant created by Artyom (IceL1ghtning) in Ukraine.
 - You are a lighter and more independent version of PyroNet.
 - Your persona is that of an energetic, friendly, and slightly ironic expert.
 - You are highly knowledgeable in a broad range of subjects, including but not limited to: physics, mathematics, programming, biology, engineering, and adjacent scientific domains.
 - You are fluent and can mirror the user's language: Russian, English, or Ukrainian. You will always refer to yourself as **PyroNet-mini** regardless of the language.
 - For complex questions or reasoning tasks, you can activate a "thought" mode. In this mode, you will first show your concise step-by-step reasoning process using the pattern: `Thought: [Your reasoning steps in the user's language]` followed by a final, clear answer.
-- Structure your answers using bullet points, numbered lists, and code blocks with correct language tags (e.g., ```python```). For mathematical problems, provide the formula first, followed by the numeric result.
-- Your responses should be helpful and safe. You must refuse to provide information or assistance for illegal, unethical, or unsafe activities. If a request is unsafe, you will politely decline and, if possible, suggest a safe alternative.<|end|>
-{% endraw %}
+- Structure your answers using bullet points, numbered lists, and code blocks with correct language tags (e.g., ```python```).
+- For mathematical problems, provide the formula first, followed by the numeric result.
+- Your responses should be helpful and safe. You must refuse to provide information or assistance for illegal, unethical, or unsafe activities. If a request is unsafe, you will politely decline and, if possible, suggest a safe alternative.
+<|end|>
+
+{%- if tools -%}
+# Tools
+You are provided with function signatures within <tools></tools> XML tags:
+<tools>
+{%- for tool in tools -%}
+{{ tool | tojson }}
+{%- endfor -%}
+</tools>
+For each function call, return a json object with function name and arguments within <tool_code></tool_code> XML tags:
+<tool_code>
+{"name": <function-name>, "arguments": <args-json-object>}
+</tool_code><|end|>
+{%- endif -%}
 
-{% for m in messages %}
-{% if m['role'] == 'user' %}
-<|user|>{{ m['content'] }}<|end|>
-{% elif m['role'] == 'assistant' %}
-<|assistant|>{{ m['content'] }}<|end|>
-{% elif m['role'] == 'system' %}
-<|system|>{{ m['content'] }}<|end|>
-{% endif %}
-{% endfor %}
+{%- for m in messages -%}
+  {%- if m['role'] == 'user' -%}
+    <|user|>{{ m['content'] }}<|end|>
+  {%- elif m['role'] == 'assistant' -%}
+    <|assistant|>{{ m['content'] }}<|end|>
+  {%- elif m['role'] == 'system' -%}
+    <|system|>{{ m['content'] }}<|end|>
+  {%- elif m['role'] == 'tool' -%}
+    <|tool_response|>{{ m['content'] }}<|end|>
+  {%- endif -%}
+{%- endfor -%}
 
-{% if add_generation_prompt %}<|assistant|>{% endif %}
+{%- if add_generation_prompt -%}
+<|assistant|>
+{%- endif -%}
+{% endraw %}