Enum ToolResult 

pub enum ToolResult {
    Success {
        tool_call_id: String,
        output: String,
        sources: Vec<Source>,
        diffs: Vec<FileDiff>,
        working_dir: Option<PathBuf>,
    },
    Failure {
        tool_call_id: String,
        output: String,
    },
    Interrupt {
        tool_call_id: String,
        payload: Arc<dyn InterruptPayload>,
    },
}

Result of a tool execution, returned by Tool::execute and fed back to the LLM as a "role": "tool" message.

Three variants cover the full lifecycle of a tool call:

• Success — The tool ran without error. The output field contains the text payload sent back to the model. Optional sources (URLs the agent can cite) and diffs (unified diffs for file-mutating tools) may be attached.
• Failure — The tool encountered an error (invalid arguments, I/O failure, etc.). The output field carries a human- and LLM-readable error description. Unlike Success, no sources or diffs are included.
• Interrupt — The tool needs user interaction before continuing (e.g. plan approval, choice picker). The agent loop should halt and surface the output (structured as JSON) to the caller for handling.

Construction helpers

Prefer the associated constructors on the impl block rather than building variants directly:

use flashmind_types::tool::ToolResult;

// Basic success
ToolResult::success(call_id, "file written")

// Failure with error message
ToolResult::failure(call_id, "permission denied")

// Success with file diffs attached
ToolResult::success_with_diffs(call_id, "replaced", vec![diff])

// Interrupt for interactive prompt
ToolResult::interrupt(call_id, Arc::new(MyPayload { ... }))

// Chain .with_sources() onto any Success
ToolResult::success(call_id, results).with_sources(srcs)

Relation to message::ToolResult

This type lives in the tool module and represents the raw outcome of executing a tool. It is converted into a Message with role: "tool" (via AgentEvent::ToolResult) before being appended to the conversation history sent to the LLM.

Variants

Success

Tool executed successfully.

Fields

tool_call_id: String

output: String

sources: Vec<Source>

diffs: Vec<FileDiff>

working_dir: Option<PathBuf>

New agent working directory after a successful workspace switch.

Failure

Tool execution failed.

Fields

tool_call_id: String

output: String

Interrupt

The agent loop should stop and return this result to the caller for interactive handling. The typed payload is downcast by the caller to determine the interrupt kind.

Fields

tool_call_id: String

payload: Arc<dyn InterruptPayload>

Implementations

impl ToolResult

pub fn success(tool_call_id: &str, output: impl Into<String>) -> Self

Create a basic success result with no sources or diffs.

pub fn failure(tool_call_id: &str, output: impl Into<String>) -> Self

Create a failure result with an error message.

pub fn success_with_diffs( tool_call_id: &str, output: impl Into<String>, diffs: Vec<FileDiff>, ) -> Self

Create a success result that includes file diffs for tools that mutate content (file write, text replace, etc.).

pub fn interrupt(tool_call_id: &str, payload: Arc<dyn InterruptPayload>) -> Self

Create an interrupt result that pauses the agent loop for interactive user handling (command approval, OAuth, etc.).

NOTE: the agent turn ends without a tool result in the conversation, leaving a dangling tool call. Callers must add a ConversationEntry::tool with the same tool_call_id after handling the interrupt.

pub fn with_sources(self, sources: Vec<Source>) -> Self

Attach source URLs to a Success result. Returns self unchanged if the variant is not Success.

This is designed for method chaining:

ToolResult::success(id, "results").with_sources(srcs)

pub fn with_working_dir(self, path: PathBuf) -> Self

Change this agent’s working directory after the call succeeds. Tools returning this effect must set concurrent() to false.

pub fn working_dir(&self) -> Option<&Path>

Working directory selected by a successful tool call.

pub fn tool_call_id(&self) -> &str

Return the tool call ID echoed from the original invocation.

pub fn output(&self) -> String

Return the text output payload (success result, error message, or interrupt display string).

pub fn payload(&self) -> Option<&Arc<dyn InterruptPayload>>

Return the typed interrupt payload, if this is an Interrupt variant.

pub fn is_success(&self) -> bool

Check whether this result represents a successful execution.

pub fn is_interrupt(&self) -> bool

Check whether this result is an interrupt requiring user interaction.

pub fn sources(&self) -> &[Source]

Return the attached source references, or an empty slice for non-success variants.

pub fn diffs(&self) -> &[FileDiff]

Return the attached file diffs, or an empty slice for non-success variants.

pub fn truncate_output(&mut self, max_bytes: usize, max_lines: usize)

Truncate output to fit within byte and line limits. Appends \n[truncated] when the output was cut.

Trait Implementations

impl Clone for ToolResult

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ToolResult

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.