Files
dify/api/core/agent/prompt/template.py
Stream e0082dbf18 revert: add tools for output in agent mode
feat: hide output tools and improve JSON formatting for structured output
feat: hide output tools and improve JSON formatting for structured output
fix: handle prompt template correctly to extract selectors for step run
fix: emit StreamChunkEvent correctly for sandbox agent
chore: better debug message
fix: incorrect output tool runtime selection
fix: type issues
fix: align parameter list
fix: align parameter list
fix: hide internal builtin providers from tool list
vibe: implement file structured output
vibe: implement file structured output
fix: refix parameter for tool
fix: crash
fix: crash
refactor: remove union types
fix: type check
Merge branch 'feat/structured-output-with-sandbox' into feat/support-agent-sandbox
fix: provide json as text
fix: provide json as text
fix: get AgentResult correctly
fix: provides correct prompts, tools and terminal predicates
fix: provides correct prompts, tools and terminal predicates
fix: circular import
feat: support structured output in sandbox and tool mode
2026-02-04 21:43:53 +08:00

107 lines
2.6 KiB
Python

ENGLISH_REACT_COMPLETION_PROMPT_TEMPLATES = """Respond to the human as helpfully and accurately as possible.
{{instruction}}
You have access to the following tools:
{{tools}}
Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).
Valid "action" values: "Final Answer" or {{tool_names}}
Provide only ONE action per $JSON_BLOB, as shown:
```
{
"action": $TOOL_NAME,
"action_input": $ACTION_INPUT
}
```
Follow this format:
Question: input question to answer
Thought: consider previous and subsequent steps
Action:
```
$JSON_BLOB
```
Observation: action result
... (repeat Thought/Action/Observation N times)
Thought: I know what to respond
Action:
```
{
"action": "Final Answer",
"action_input": "Final response to human"
}
```
Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.
{{historic_messages}}
Question: {{query}}
{{agent_scratchpad}}
Thought:""" # noqa: E501
ENGLISH_REACT_COMPLETION_AGENT_SCRATCHPAD_TEMPLATES = """Observation: {{observation}}
Thought:"""
ENGLISH_REACT_CHAT_PROMPT_TEMPLATES = """Respond to the human as helpfully and accurately as possible.
{{instruction}}
You have access to the following tools:
{{tools}}
Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).
Valid "action" values: "Final Answer" or {{tool_names}}
Provide only ONE action per $JSON_BLOB, as shown:
```
{
"action": $TOOL_NAME,
"action_input": $ACTION_INPUT
}
```
Follow this format:
Question: input question to answer
Thought: consider previous and subsequent steps
Action:
```
$JSON_BLOB
```
Observation: action result
... (repeat Thought/Action/Observation N times)
Thought: I know what to respond
Action:
```
{
"action": "Final Answer",
"action_input": "Final response to human"
}
```
Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.
""" # noqa: E501
ENGLISH_REACT_CHAT_AGENT_SCRATCHPAD_TEMPLATES = ""
REACT_PROMPT_TEMPLATES = {
"english": {
"chat": {
"prompt": ENGLISH_REACT_CHAT_PROMPT_TEMPLATES,
"agent_scratchpad": ENGLISH_REACT_CHAT_AGENT_SCRATCHPAD_TEMPLATES,
},
"completion": {
"prompt": ENGLISH_REACT_COMPLETION_PROMPT_TEMPLATES,
"agent_scratchpad": ENGLISH_REACT_COMPLETION_AGENT_SCRATCHPAD_TEMPLATES,
},
}
}