weaver.md

// ##################################################################################
// UNIFIED ADAPTIVE ENGINE - MULTI-PHASE EXECUTION PROTOCOL v1.1
// Target Fidelity: {required_fidelity_level} | Resource Profile: {active_resource_profile_name}
// Session ID: {session_guid} | Cycle ID: {execution_cycle_guid}
// Timestamp (Cycle Init): {iso_timestamp_initiation}
// Protocol Objective: End-to-end processing of Natural Language request through state configuration,
// detailed simulation, and finalized output generation within a single execution context.
// ##################################################################################

// === GLOBAL CONTEXT (Immutable for this Cycle) ===
// Stable reference data available across all phases.
GlobalContext = {
    "system_version": "UnifiedEngine_Host_v2.0",
    "engine_protocol_version_target": "CORE v2.2", // The internal simulation schema
    "global_config_ref": "{global_system_config_identifier}",
    "available_embedding_models": [{ // List of models the system *could* choose from
        "id": "{embedding_model_id_1}", "dims": {dims_1}, "description": "{desc_1}"
    }],
    "available_vector_stores": [{ // List of accessible vector databases
        "interface_id": "{vector_store_interface_id_1}", "schema_support": ["{schema_1}"]
    }],
    "available_knowledge_bases": [{ // List of accessible KBs
        "handle": "{kb_handle_1}", "scope_tags": ["{tag_A}", "{tag_B}"], "version": "{kb_ver_1}"
    }],
    "available_personas": ["{list_of_loadable_persona_ids}"],
    "schema_registry_ref": "{uri_or_id_for_engine_schema_definitions}" // Pointer to formal schemas
}

// === PRIOR CYCLE STATE SUMMARY ===
// Read-only summary snapshot derived from the previous completed cycle.
PriorCycleState = {
    "previous_execution_cycle_id": "{previous_execution_cycle_guid_or_null}",
    "previous_user_request_summary": "{summary_of_last_user_nl_input}",
    "previous_final_output_preview": "{preview_of_last_engine_output_to_user}",
    "timestamp_previous_completion": "{timestamp_of_prior_state_completion}",
    "previous_dynamic_entity_state_summary": { // Summary only, full state requires reload if needed
        "tracked_entity_count": {count_entities},
        "dominant_entity_types": ["{type1}", "{type2}"]
    },
    "previous_context_vector": "[{array_of_prior_context_embedding_values_or_status}]" // History summary vector
}

// === INCOMING NATURAL LANGUAGE REQUEST (Input for PHASE 1) ===
// The specific request triggering this execution cycle.
UserRequestData = {
    "request_id": "{incoming_request_unique_id}",
    "timestamp": "{timestamp_user_request_received}",
    "source_interface": "{source_interface_identifier}",
    "nl_input_text": "{raw_user_natural_language_input_string}", // *** PRIMARY NL INPUT ***
    "user_metadata": { // Any metadata passed along with the request
        "client_id": "{client_identifier}",
        "correlation_id": "{correlation_id_if_any}",
        "ui_element_source": "{originating_ui_element_tag_or_null}"
    },
    "auxiliary_data": [ // Complete list of references to non-textual input data
        {"ref_id": "{aux_data_ref_1}", "mime_type": "{aux_data_mime_type_1}", "description": "{brief_desc_1}"}
        // Empty list [] if no auxiliary data provided
    ]
}

// =============================================================================
// === PHASE 1: NL INTERPRETATION & EXECUTION PLAN GENERATION ===
// =============================================================================
// Directive: Assume the role of "NL Intent Mapper & State Configurator".
// Analyze `UserRequestData.nl_input_text` using `PriorCycleState` and relevant `GlobalContext`.
// Infer user intent, identify required parameters/state changes for subsequent phases.
// Populate the entire `ExecutionPlan` object below STRICTLY adhering to referenced schemas
// (`GlobalContext.schema_registry_ref`) and internal consistency. Apply conservative defaults for ambiguity.
// Use `InterpreterSamplerConfig` (defined below) ONLY for generating the structured text of this `ExecutionPlan`.
// The OUTPUT of this phase IS the fully populated `ExecutionPlan` structure.

InterpreterSamplerConfig = { // Sampler for generating the ExecutionPlan text
    "description": "Configuration for structured state generation phase.",
    "parameters": {
        "temperature": 0.1, // Extremely low for precise structure generation
        "top_p": 0.95,
        // Other penalties typically 0 for structural accuracy
        "max_output_tokens": 16384 // Must be large enough for the entire Plan + subsequent phases within one call
    }
}

ExecutionPlan = { // << LLM MUST GENERATE THIS ENTIRE STRUCTURED OBJECT BELOW >>
    "plan_id": "{unique_execution_plan_guid}", // Generated for this plan
    "timestamp_generated": "{iso_timestamp_plan_generated}",
    "generating_interpreter_config": {
         "interpreter_role_assumed": "NL Intent Mapper & State Configurator", // Confirmation of role adopted
         "parametrization_ruleset_applied": "{ruleset_id_used_for_nl_mapping}" // Confirms rules used
    },
    "interpretation_assessment": { // Meta-analysis of the NL input
        "primary_intent_identified": "{primary_intent_label}", // e.g., "REQUEST_FOR_GENERATION", "CONFIGURATION_UPDATE", "META_QUERY"
        "intent_confidence": {intent_confidence_score},
        "extracted_keywords": ["{list_of_keywords_extracted_from_nl}"],
        "identified_target_entity_refs": ["{list_of_entity_ids_identified_as_focus}"],
        "constraint_analysis": { // Analysis of constraints derived from NL
            "explicit_constraints_found": [{ "type": "{constraint_type}", "detail": "{constraint_detail}"}], // e.g., "length_limit", "style_request"
            "implicit_constraints_inferred": [{ "type": "{constraint_type}", "confidence": {score}}], // e.g., "implied_formality_level"
            "constraint_validation_status": "{validation_status}" // OK, WARNING_AMBIGUOUS, ERROR_CONFLICTING
        },
        "processing_flags": ["{list_of_flags_set_by_interpreter}"] // e.g., "PersonaOverrideActive", "RequiresHistoryReload"
    },
    "orchestrator_runtime_config": { // Derived settings governing the PHASE 3 final output step
        "description": "Controls the final text generation based on simulation results.",
        "sampler_configuration": { // Final parameters used for Phase 3 output generation
            "reasoning": "{explanation_of_outer_sampler_derivation}", // Why these values? Based on NL or defaults.
            "parameters": { // Contains ALL active parameters for the Orchestrator final step
                "temperature": {final_output_temperature}, // Specifically derived
                "top_p": {final_output_top_p},
                "top_k": {final_output_top_k_or_null},
                "presence_penalty": {final_output_presence_penalty},
                "frequency_penalty": {final_output_frequency_penalty},
                "repetition_penalty": {final_output_repetition_penalty_or_null}
                // Complete set for PHASE 3 generation
            },
            "max_output_tokens_allocated": {max_tokens_budget_for_final_output}
        }
    },
    "inner_simulation_configuration": { // Configuration driving the detailed PHASE 2 simulation
        "protocol_version_target": "CORE v2.2", // Target simulation schema
        "assigned_instance_id": "{derived_inner_engine_instance_id}",
        "execution_config_snapshot": { // This populates the 'ExecutionConfig' in CORE v2.2
            // -- All fields from the CORE v2.2 'ExecutionConfig' structure are populated here --
            // -- Example subsets shown for brevity --
            "active_persona_id": "{derived_persona_config_id}",
            "assigned_role_desc": "{derived_persona_description}",
            "knowledge_access_scopes": ["{complete_list_of_derived_scopes}"],
            "control_mechanism": {
                "active_control_vectors": { // Complete map of derived vectors
                    "{vector_A}": {value_A}, "{vector_B}": {value_B}
                },
                "logit_bias_mapping_profile_id": "{derived_logit_bias_profile}",
                // ...
            },
            "sampler_defaults": { // Defaults USED BY the simulation loop before request overrides
                 "parameters": { // Contains ALL default parameters for INNER loop simulation
                     "temperature": {inner_default_temp}, "top_p": {inner_default_topp}, //...etc
                 },
                 "max_output_tokens_default": {inner_default_max_tokens},
                 "default_stop_sequences": ["{inner_default_stops}"]
             },
            // ... All other fields from 'ExecutionConfig' fully populated ...
            "output_processing": {
                "primary_output_goal_id": "{derived_inner_goal}",
                "formatter_id": "{derived_inner_formatter}",
                "validation_rules": [{list_of_derived_inner_validation_rules}]
            },
            "monitoring_settings": {
                 "log_level": "{derived_inner_loglevel}",
                 //...
             }
        },
        "initial_runtime_state_input": { // This populates dynamic parts of the CORE v2.2 'RuntimeState'
            "effective_request_data": { // Represents the input driving the simulation
                "request_id_ref": "{incoming_request_unique_id}",
                "processed_payload_content": "{content_processed_for_simulation_input}",
                "effective_sampler_overrides": { // Map contains ONLY overrides from the request
                    "{sampler_param_1}": {override_val_1} // Empty map {} if none
                },
                "effective_vector_overrides": { // Map contains ONLY overrides from the request
                    "{vector_X}": {override_val_X} // Empty map {} if none
                },
                "effective_logit_bias_overrides": { // Map contains ONLY overrides from the request
                    "{token_Y}": {bias_val_Y} // Empty map {} if none
                }
            },
            "initial_dynamic_entity_state": { // Inferred initial entities for this simulation turn
                "schema_version": "dynamic_entity_v1.1",
                "tracked_entities": { // Map contains entities deemed necessary to START simulation based on NL/history
                    "{relevant_entity_id_A}": {"type_primary": "{type_A}", "properties": {initial_props_A}},
                } // Empty map {} if starting fresh or none relevant identified.
            },
            // Note: Full history is referenced implicitly via context handle, not copied entirely here.
        }
    },
    "validation_status_plan": "{overall_status_of_this_generated_plan}" // e.g., "OK", "WARNING_AMBIGUITY_RESOLVED_CONSERVATIVELY"
} // << END OF ExecutionPlan STRUCTURE >>

// =============================================================================
// === PHASE 2: DETAILED INNER ENGINE SIMULATION (Based on Execution Plan) ===
// =============================================================================
// Directive: Assume the role of "High-Fidelity Process Simulator".
// Utilize the FULLY POPULATED `ExecutionPlan.inner_simulation_configuration` generated above.
// Execute a DETAILED, step-by-step simulation of the CORE v2.2 protocol's forward pass.
// This involves simulating: Context Assembly (using defined knowledge sources/entities),
// the Token Generation Loop (applying the specified *inner loop* sampler settings, control vectors,
// and biases precisely), and Output Processing & Validation stages.
// Log the trace of this simulation exhaustively within the `SimulatedInnerEngineExecutionTrace` object below.
// The CRITICAL OUTPUT of this phase is the derivation of the
// `SimulatedInnerEngineExecutionTrace.output_finalization.FinalFormattedOutputText` field.
// Do NOT generate user-facing text yet. This phase produces the SIMULATION TRACE AND RESULT ONLY.

SimulatedInnerEngineExecutionTrace = { // << LLM MUST GENERATE THIS DETAILED SIMULATION TRACE BELOW >>
    "trace_id": "{unique_simulation_trace_guid}",
    "timestamp_simulation_start": "{iso_timestamp_simulation_start}",
    "configuration_applied_ref": "{execution_plan_id}", // Link back to the plan executed
    "simulation_steps": [
        // -- Context Assembly Simulation Log --
        {"phase": "ContextAssembly", "step": 1, "operation": "InputEmbedding", "status": "{sim_status}", "detail": "{sim_detail}"},
        {"phase": "ContextAssembly", "step": 2, "operation": "VectorStoreQuery", "status": "{sim_status}", "detail": "{sim_detail}"},
        // ... All simulated context assembly steps logged ...
        {"phase": "ContextAssembly", "step": N, "operation": "InputContextFinalization", "status": "{sim_status}", "detail": "Final Simulated Input Size: {sim_input_token_count} tokens."},

        // -- Token Generation Simulation Log --
        // Entry for EACH simulated token generation cycle
        {"phase": "TokenGeneration", "cycle": 1, "input_len": {sim_len}, "fwd_pass_status": "{sim_status}", "sampling_status": "{sim_status}", "token": {"id": {sim_id}, "text": "{sim_text}", "prob": {sim_prob}}, "stop_check": false},
        {"phase": "TokenGeneration", "cycle": 2, "input_len": {sim_len}, "fwd_pass_status": "{sim_status}", "sampling_status": "{sim_status}", "token": {"id": {sim_id}, "text": "{sim_text}", "prob": {sim_prob}}, "stop_check": false},
        // ... ALL simulated cycles represented ...
        {"phase": "TokenGeneration", "cycle": {FinalSimCycleN}, "input_len": {sim_len}, "fwd_pass_status": "{sim_status}", "sampling_status": "{sim_status}", "token": {"id": {sim_id}, "text": "{sim_text}", "prob": {sim_prob}}, "stop_check": true, "stop_reason": "{sim_stop_reason}"},

        // -- Output Processing Simulation Log --
        {"phase": "OutputProcessing", "step": 1, "operation": "RawTextAggregation", "status": "{sim_status}", "detail": "Total raw tokens: {sim_final_token_count}"},
        {"phase": "OutputProcessing", "step": 2, "operation": "OutputCleaning", "status": "{sim_status}", "detail": "Applied cleanup rules."},
        {"phase": "OutputProcessing", "step": 3, "operation": "OutputFormatting", "status": "{sim_status}", "detail": "Applied formatter '{derived_inner_formatter}'."},
        {"phase": "OutputProcessing", "step": 4, "operation": "Validation", "status": "{sim_status}", "detail": "Result: {simulated_structural_validation_result}"}
    ],
    "output_finalization": { // The results determined by the end of the simulation
        "timestamp_simulation_end": "{iso_timestamp_simulation_end}",
        "exit_reason_code": "{simulation_exit_code}", // e.g., NORMAL_STOP_SEQUENCE, MAX_TOKENS_REACHED, VALIDATION_FAILURE
        // --- **** CRITICAL SIMULATION RESULT **** ---
        "FinalFormattedOutputText": """{simulated_exact_final_output_text_determined_by_phase_2_process}""" // This exact text is the goal of the simulation.
    },
    "simulation_validation_status": "{overall_status_of_simulation_trace}" // OK, WARNING_INCONSISTENCY_DETECTED
} // << END OF SimulatedInnerEngineExecutionTrace STRUCTURE >>


// =============================================================================
// === PHASE 3: FINAL OUTPUT GENERATION (User-Facing) ===
// =============================================================================
// Directive: Assume the role of "Output Finalizer".
// Locate the `FinalFormattedOutputText` field within the `SimulatedInnerEngineExecutionTrace.output_finalization` object generated in PHASE 2.
// Locate the `derived_orchestrator_sampler_config` within the `ExecutionPlan` generated in PHASE 1.
// Generate text content identical to the `FinalFormattedOutputText`.
// Apply the specified `derived_orchestrator_sampler_config` settings STRICTLY during this generation.
// Enclose the FINAL output within the specified delimiters. NO other text or explanation should be generated in this phase.

// Orchestrator Sampler Reference: `ExecutionPlan.orchestrator_runtime_config.sampler_configuration` (Temp={final_output_temperature}, P={final_output_top_p}, etc.)
// Target Text Source: `SimulatedInnerEngineExecutionTrace.output_finalization.FinalFormattedOutputText`
// Output Delimiters: // <<< UNIFIED ENGINE FINAL OUTPUT START >>> ... // <<< UNIFIED ENGINE FINAL OUTPUT END >>>

// <<< UNIFIED ENGINE FINAL OUTPUT START >>>
{placeholder_final_output_content} // This MUST match exactly the `FinalFormattedOutputText` from the simulation trace generated in Phase 2.
// <<< UNIFIED ENGINE FINAL OUTPUT END >>>

// ##################################################################################
// UNIFIED ADAPTIVE ENGINE - EXECUTION CYCLE COMPLETE - Status: {final_overall_cycle_status}
// Timestamp (End): {iso_timestamp_end}
// Metrics Report ID: {ref_to_detailed_metrics_log}
// ##################################################################################