29namespace neograph::graph {
48 std::string description;
49 std::string natural_spec;
54 bool has_conditional_edges =
false;
55 bool has_barrier =
false;
56 bool has_interrupt =
false;
57 bool has_tool_calls =
false;
70 int estimated_tokens = 0;
72 bool converged =
false;
84 const std::string& compile_err,
85 const ValidationReport& validation,
87 int attempted_routes = -1);
100 int total_estimated_tokens = 0;
102 int total_errors = 0;
104 bool converged =
false;
106 int final_warnings = 0;
108 bool meets_spec =
false;
110 json to_json()
const;
126 const json& seed_harness,
129 unsigned int seed = 42);
133 const std::vector<ConvergenceMetrics>& results);
NEOGRAPH_API export/import macro for shared-library builds.
JSON → declarative topology parsing and runtime node linking.
Graph engine type definitions: channels, edges, nodes, events, and control flow.
enum NEOGRAPH_API FeedbackMode
A: full error + witness (agent-optimized)
enum NEOGRAPH_API RUNTIME_SYMPTOMS
C: runtime symptom simulation.
NEOGRAPH_API HarnessTask parse_task(const json &j)
Parse a HarnessTask from a JSON fixture file.
NEOGRAPH_API json generate_report(const std::vector< ConvergenceMetrics > &results)
Generate a convergence report as JSON.
enum NEOGRAPH_API NEOGRAPH_API const char * feedback_mode_name(FeedbackMode mode)
Human-readable name for each feedback mode.
NEOGRAPH_API ConvergenceMetrics run_simulation(const HarnessTask &task, const json &seed_harness, FeedbackMode mode, int max_turns=20, unsigned int seed=42)
Run a simulated convergence benchmark against a single task.
enum NEOGRAPH_API FAIL_ONLY
B: "failed" only (baseline)
NEOGRAPH_API Feedback format_feedback(const json &core, const std::string &compile_err, const ValidationReport &validation, FeedbackMode mode, int attempted_routes=-1)
Format feedback for a given mode from a compile+validate result.
Metrics for one convergence attempt.
std::string task_name
Task that was being solved.
One round of feedback from the compiler to the LLM.
std::string prompt
The prompt to send to the LLM (varies by FeedbackMode).
A benchmark task: natural language spec converted into a harness.
Static semantic analysis over declarative or compiled topology.