MCPcopy Create free account
hub / github.com/FastMAS/KVCOMM / _process_inputs

Method _process_inputs

KVCOMM/agents/math_solver.py:30–122  ·  view source on GitHub ↗

Prepare prompts, possibly set anchors, and return mode hints.

(
        self,
        raw_inputs:Dict[str,str],
        spatial_info:Dict[str,Dict],
        temporal_info:Dict[str,Dict],
        mode: str = "default",
        **kwargs,
    )

Source from the content-addressed store, hash-verified

28 self.constraint = self.prompt_set.get_constraint(self.role)
29
30 async def _process_inputs(
31 self,
32 raw_inputs:Dict[str,str],
33 spatial_info:Dict[str,Dict],
34 temporal_info:Dict[str,Dict],
35 mode: str = "default",
36 **kwargs,
37 )->Dict[str, Any]:
38 """Prepare prompts, possibly set anchors, and return mode hints."""
39 if mode == "allow_kv_reuse":
40 request_uid = raw_inputs.get("_request_uid") or kwargs.get("request_uid")
41 if request_uid is None:
42 raise ValueError("request_uid is required for request-scoped anchor updates.")
43
44 preferred_mode = "kv_reuse"
45 agent_memory = self.llm._ensure_agent_memory(self.id)
46
47 if self.llm.has_prefix_initialized(self.id) and "placeholder_info" in agent_memory:
48 for agent_id, info in spatial_info.items():
49 if info["role"] != "Programming Expert":
50 continue
51 answer = execute_code_get_return(
52 info["output"].lstrip("```python\n").rstrip("\n```")
53 )
54 self.llm.update_condition_anchor(
55 request_uid=request_uid,
56 owner_agent_id=agent_id,
57 message=raw_inputs["task"],
58 content=f"The answer is:\n{answer}",
59 prefix_text="The answer is:\n",
60 )
61
62 prefix_text = kwargs.get("prefix", "Q:")
63 user_content = prefix_text + raw_inputs["task"]
64 preferred_mode = self.llm.update_input_anchor(
65 request_uid=request_uid,
66 agent_id=self.id,
67 message=raw_inputs["task"],
68 user_content=user_content,
69 prefix_text=prefix_text,
70 )
71 logger.opt(colors=True).info(
72 "<green>[MODE]</green> Task: {} Agent {} ({}) mode: {}",
73 raw_inputs["task"],
74 self.id,
75 self.role,
76 preferred_mode,
77 )
78 return {"preferred_mode": preferred_mode, "early_response": None}
79
80 system_prompt = f"{self.constraint}"
81 user_input = "{user_question}"
82 user_prompt = self.prompt_set.get_answer_prompt(question=user_input, role=self.role)
83 spatial_str = ""
84 temporal_str = ""
85 for id, info in spatial_info.items():
86 agent_output = info["output"] if len(info["output"]) > 0 else "{agent_" + id + "_current}"
87 if info["role"] == "Programming Expert":

Callers 2

_executeMethod · 0.95
_async_executeMethod · 0.95

Calls 9

execute_code_get_returnFunction · 0.90
_ensure_agent_memoryMethod · 0.80
itemsMethod · 0.80
update_input_anchorMethod · 0.80
getMethod · 0.45
get_answer_promptMethod · 0.45

Tested by

no test coverage detected