Struct PipelineState 

pub struct PipelineState {

    pub phase: PipelinePhase,
    pub previous_phase: Option<PipelinePhase>,
    pub iteration: u32,
    pub total_iterations: u32,
    pub reviewer_pass: u32,
    pub total_reviewer_passes: u32,
    pub review_issues_found: bool,
    pub planning_prompt_prepared_iteration: Option<u32>,
    pub planning_xml_cleaned_iteration: Option<u32>,
    pub planning_agent_invoked_iteration: Option<u32>,
    pub planning_xml_extracted_iteration: Option<u32>,
    pub planning_validated_outcome: Option<PlanningValidatedOutcome>,
    pub planning_markdown_written_iteration: Option<u32>,
    pub planning_xml_archived_iteration: Option<u32>,
    pub development_context_prepared_iteration: Option<u32>,
    pub development_prompt_prepared_iteration: Option<u32>,
    pub development_xml_cleaned_iteration: Option<u32>,
    pub development_agent_invoked_iteration: Option<u32>,
    pub analysis_agent_invoked_iteration: Option<u32>,
    pub development_xml_extracted_iteration: Option<u32>,
    pub development_validated_outcome: Option<DevelopmentValidatedOutcome>,
    pub development_xml_archived_iteration: Option<u32>,
    pub review_context_prepared_pass: Option<u32>,
    pub review_prompt_prepared_pass: Option<u32>,
    pub review_issues_xml_cleaned_pass: Option<u32>,
    pub review_agent_invoked_pass: Option<u32>,
    pub review_issues_xml_extracted_pass: Option<u32>,
    pub review_validated_outcome: Option<ReviewValidatedOutcome>,
    pub review_issues_markdown_written_pass: Option<u32>,
    pub review_issue_snippets_extracted_pass: Option<u32>,
    pub review_issues_xml_archived_pass: Option<u32>,
    pub fix_prompt_prepared_pass: Option<u32>,
    pub fix_result_xml_cleaned_pass: Option<u32>,
    pub fix_agent_invoked_pass: Option<u32>,
    pub fix_result_xml_extracted_pass: Option<u32>,
    pub fix_validated_outcome: Option<FixValidatedOutcome>,
    pub fix_result_xml_archived_pass: Option<u32>,
    pub commit_prompt_prepared: bool,
    pub commit_diff_prepared: bool,
    pub commit_diff_empty: bool,
    pub commit_diff_content_id_sha256: Option<String>,
    pub commit_agent_invoked: bool,
    pub commit_xml_cleaned: bool,
    pub commit_xml_extracted: bool,
    pub commit_validated_outcome: Option<CommitValidatedOutcome>,
    pub commit_xml_archived: bool,
    pub context_cleaned: bool,
    pub agent_chain: AgentChainState,
    pub rebase: RebaseState,
    pub commit: CommitState,
    pub execution_history: Vec<ExecutionStep>,
    pub checkpoint_saved_count: u32,
    pub continuation: ContinuationState,
    pub metrics: RunMetrics,
    pub dev_fix_triggered: bool,
    pub prompt_inputs: PromptInputsState,
}

Immutable pipeline state - the single source of truth for pipeline progress.

This struct captures complete execution context and doubles as the checkpoint data structure for resume functionality. Serialize it to JSON to save state; deserialize to resume interrupted runs.

Invariants

• iteration is always <= total_iterations
• reviewer_pass is always <= total_reviewer_passes
• agent_chain maintains fallback order and retry counts
• State transitions only occur through the reduce function

See Also

• reduce for state transitions
• determine_next_effect for effect derivation

Fields

phase: PipelinePhase

previous_phase: Option<PipelinePhase>

iteration: u32

total_iterations: u32

reviewer_pass: u32

total_reviewer_passes: u32

review_issues_found: bool

planning_prompt_prepared_iteration: Option<u32>

Tracks whether the planning prompt was prepared for the current iteration.

planning_xml_cleaned_iteration: Option<u32>

Tracks whether .agent/tmp/plan.xml was cleaned for the current iteration.

planning_agent_invoked_iteration: Option<u32>

Tracks whether the planning agent was invoked for the current iteration.

planning_xml_extracted_iteration: Option<u32>

Tracks whether .agent/tmp/plan.xml was successfully extracted for the iteration.

planning_validated_outcome: Option<PlanningValidatedOutcome>

Stores the validated outcome for the current planning iteration.

planning_markdown_written_iteration: Option<u32>

Tracks whether PLAN.md has been written for the current iteration.

planning_xml_archived_iteration: Option<u32>

Tracks whether .agent/tmp/plan.xml was archived for the current iteration.

development_context_prepared_iteration: Option<u32>

Tracks whether development context was prepared for the current iteration.

Used to sequence single-task development effects.

development_prompt_prepared_iteration: Option<u32>

Tracks whether the development prompt was prepared for the current iteration.

development_xml_cleaned_iteration: Option<u32>

Tracks whether .agent/tmp/development_result.xml was cleaned for the current iteration.

development_agent_invoked_iteration: Option<u32>

Tracks whether the developer agent was invoked for the current iteration.

analysis_agent_invoked_iteration: Option<u32>

Tracks whether the analysis agent was invoked for the current iteration.

Analysis agent runs after EVERY development iteration to produce an objective assessment of progress by comparing git diff against PLAN.md. This ensures continuous verification throughout the development phase.

development_xml_extracted_iteration: Option<u32>

Tracks whether .agent/tmp/development_result.xml was extracted for the current iteration.

development_validated_outcome: Option<DevelopmentValidatedOutcome>

Stores the validated development outcome for the current iteration.

development_xml_archived_iteration: Option<u32>

Tracks whether the development XML was archived for the current iteration.

review_context_prepared_pass: Option<u32>

Tracks whether review context was prepared for the current pass.

Used to sequence single-task review effects (PrepareReviewContext -> …).

review_prompt_prepared_pass: Option<u32>

Tracks whether the review prompt was prepared for the current pass.

review_issues_xml_cleaned_pass: Option<u32>

Tracks whether .agent/tmp/issues.xml was cleaned for the current pass.

review_agent_invoked_pass: Option<u32>

Tracks whether the reviewer agent was invoked for the current pass.

review_issues_xml_extracted_pass: Option<u32>

Tracks whether .agent/tmp/issues.xml was successfully extracted for the current pass.

review_validated_outcome: Option<ReviewValidatedOutcome>

Stores the validated outcome for the current review pass.

This is used to sequence post-validation single-task effects (write markdown, archive XML) before the reducer advances to the next pass/phase.

review_issues_markdown_written_pass: Option<u32>

Tracks whether ISSUES.md has been written for the current pass.

review_issue_snippets_extracted_pass: Option<u32>

Tracks whether review issue snippets were extracted for the current pass.

review_issues_xml_archived_pass: Option<u32>

fix_prompt_prepared_pass: Option<u32>

fix_result_xml_cleaned_pass: Option<u32>

fix_agent_invoked_pass: Option<u32>

fix_result_xml_extracted_pass: Option<u32>

fix_validated_outcome: Option<FixValidatedOutcome>

fix_result_xml_archived_pass: Option<u32>

commit_prompt_prepared: bool

Tracks whether the commit prompt was prepared for the current commit attempt.

commit_diff_prepared: bool

Tracks whether the commit diff has been computed for the current attempt.

commit_diff_empty: bool

Tracks whether the computed commit diff was empty.

commit_diff_content_id_sha256: Option<String>

Content identifier (sha256 hex) of the prepared commit diff.

This is recorded when the diff is prepared and is used by orchestration guards to avoid reusing stale materialized prompt inputs across checkpoint resumes or when tmp artifacts change.

commit_agent_invoked: bool

Tracks whether the commit agent was invoked for the current commit attempt.

commit_xml_cleaned: bool

Tracks whether .agent/tmp/commit_message.xml was cleaned for the current attempt.

commit_xml_extracted: bool

Tracks whether .agent/tmp/commit_message.xml was extracted for the current attempt.

commit_validated_outcome: Option<CommitValidatedOutcome>

Stores the validated commit outcome for the current attempt.

commit_xml_archived: bool

Tracks whether commit XML has been archived for the current attempt.

context_cleaned: bool

agent_chain: AgentChainState

rebase: RebaseState

commit: CommitState

execution_history: Vec<ExecutionStep>

checkpoint_saved_count: u32

Count of CheckpointSaved events applied to state.

This is a reducer-visible record of checkpoint saves, intended for observability and tests that enforce checkpointing happens via effects.

continuation: ContinuationState

Continuation state for development iterations.

Tracks context from previous attempts when status is “partial” or “failed” to enable continuation-aware prompting.

metrics: RunMetrics

Run-level execution metrics.

This is the single source of truth for all iteration/attempt/retry/fallback statistics. Updated deterministically by the reducer based on events.

dev_fix_triggered: bool

Whether TriggerDevFixFlow has been executed in the current AwaitingDevFix phase.

This flag is set to true when DevFixTriggered event is reduced. It ensures the event loop allows at least one iteration to execute TriggerDevFixFlow before checking completion, preventing premature exit when max iterations is imminent.

prompt_inputs: PromptInputsState

Canonical, reducer-visible prompt inputs after oversize materialization.

This is the single source of truth for any inline-vs-reference and model-budget truncation decisions. Effects must not silently re-truncate or re-reference content on retries; instead, they should consume these materialized inputs (or materialize them exactly once per content id).

Implementations

impl PipelineState

pub fn initial(developer_iters: u32, reviewer_reviews: u32) -> Self

pub fn initial_with_continuation( developer_iters: u32, reviewer_reviews: u32, continuation: ContinuationState, ) -> Self

Create initial state with custom continuation limits from config.

Use this when you need to load XSD retry and continuation limits from unified config. Example:

// Config semantics: max_dev_continuations counts continuation attempts *beyond*
// the initial attempt. ContinuationState::max_continue_count semantics are
// "maximum total attempts including initial".
let continuation = ContinuationState::with_limits(
    config.general.max_xsd_retries,
    1 + config.general.max_dev_continuations,
    config.general.max_same_agent_retries,
);
let state = PipelineState::initial_with_continuation(dev_iters, reviews, continuation);

pub fn is_complete(&self) -> bool

Returns true if the pipeline is in a terminal state for event loop purposes.

Terminal States

• Complete phase: Always terminal (successful completion)
• Interrupted phase: Terminal under these conditions:
  1. A checkpoint has been saved (normal Ctrl+C interruption path)
  2. Transitioning from AwaitingDevFix phase (failure handling completed)

AwaitingDevFix → Interrupted Path

When the pipeline encounters a terminal failure (e.g., AgentChainExhausted), it transitions through AwaitingDevFix phase where:

1. TriggerDevFixFlow effect writes completion marker to filesystem
2. Dev-fix agent is dispatched (optional remediation attempt)
3. CompletionMarkerEmitted event transitions to Interrupted phase

At this point, the completion marker has been written, signaling external orchestration that the pipeline has terminated. The SaveCheckpoint effect will execute next, but the phase is already considered terminal because the failure has been properly signaled.

Edge Cases

An Interrupted phase without a checkpoint and without previous_phase context is NOT considered terminal. This can occur when resuming from a checkpoint that was interrupted mid-execution.

Non-Terminating Pipeline Architecture

The pipeline is designed to never exit early. All failure paths route through AwaitingDevFix → TriggerDevFixFlow → completion marker write → Interrupted. This ensures orchestration can reliably detect completion via the marker file, even when budget is exhausted or all agents fail.

Terminal states:

• Complete: Normal successful completion
• Interrupted with checkpoint saved: Resumable state
• Interrupted from AwaitingDevFix: Completion marker written, failure signaled

pub fn current_head(&self) -> String

Trait Implementations

impl Clone for PipelineState

fn clone(&self) -> PipelineState

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PipelineState

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for PipelineState

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl From<PipelineCheckpoint> for PipelineState

fn from(checkpoint: PipelineCheckpoint) -> Self

Converts to this type from the input type.

impl Serialize for PipelineState

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.