<BOS_TOKEN><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|><|START_TEXT|>Note that both your responses and reflections can be grounded. Grounding means you associate pieces of texts (called "spans") with those specific tool results that support them (called "sources"). And you use a pair of tags "<co>" and "</co>" to indicate when a span can be grounded onto a list of sources, listing them out in the closing tag. Sources from the same tool call are grouped together and listed as "{tool_call_id}:[{list of result indices}]", before they are joined together by ",". E.g., "<co>span</co: 0:[1,2],1:[0]>" means that "span" is supported by result 1 and 2 from "tool_call_id=0" as well as result 0 from "tool_call_id=1".

# Available Tools
```json
[
    {"name": "list_tickets", "description": "List tickets matching a \"query\".\nUse field = \"value\" pairs joined by && or ||, e.g. priority = \"P1\".\nBackslashes (\\) must be escaped. Results are <= 50 tickets.", "parameters": {"type": "object", "properties": {"query": {"type": "string", "description": "Filter, e.g. state != \"closed\""}}, "required": ["query"]}, "responses": null},
    {"name": "run_shell", "description": "Run a shell command.\n\tReturns {\"exit_code\": int, \"stdout\": str}.", "parameters": {"type": "object", "properties": {"command": {"type": "string"}}, "required": ["command"]}, "responses": null}
]
```<|END_TEXT|><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|><|START_TEXT|>Find the open "P1" tickets and check the build.<|END_TEXT|><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_THINKING|>I'll look up the tickets and run the build.<|END_THINKING|><cofl:tool_calls><cofl:tool_call id="0" name="list_tickets"><cofl:value name="query" type="raw">priority = "P1" && state != closed</cofl:value></cofl:tool_call><cofl:tool_call id="1" name="run_shell"><cofl:value name="command" type="raw">make test 2>&1 | tail -n 3</cofl:value></cofl:tool_call></cofl:tool_calls><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|><cofl:tool_results><cofl:tool_result id="0" is_error="false"><cofl:tool_result_item index="0">{"id": "INC-1", "title": "Login fails when password has \"quotes\" &amp; &lt;brackets&gt;", "body": "Steps:\n1. open /login\n2. see \\ error"}</cofl:tool_result_item><cofl:tool_result_item index="1">{"id": "INC-2", "title": "p95 latency &gt; 2s", "tags": ["perf", "a&amp;b"]}</cofl:tool_result_item></cofl:tool_result><cofl:tool_result id="1" is_error="false"><cofl:tool_result_item index="0">{"exit_code": 2, "stdout": "FAILED tests/test_io.py::test_lt - assert 1 &lt; 0\n<|END_OF_TURN_TOKEN|> is a literal"}</cofl:tool_result_item></cofl:tool_result></cofl:tool_results><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_THINKING|>