pulselake_check_request
ChatGPTRecover a delayed or disconnected tool response using its receipt_id. This checks delivery, not simulation progress. If completed, use the original returned result; if pending, wait retry_after_seconds. Never resubmit a paid action when its outcome is unknown.
pulselake_check_simulation_state
ChatGPTAuthoritatively check whether a Simulation Research request is still being configured, awaiting approval, queued/running as a real submitted job, or terminal. Use the most advanced identifier currently available: job_id after submission, approval_token after preview, otherwise guide_token. Configuration activity is never reported as execution. Do not repeatedly poll setup or approval states; ask the user for the required input or approval instead. If the guide reference is lost or rejected, call with no identifiers to find this user's saved configurations. No credits are used.
pulselake_continue_simulation
ChatGPTValidate and save answers to the guide's typed configuration questions, such as concept or experience, decision objective, audience, panel design, questions, criteria, source, and report settings. Returns remaining questions or, when configuration is complete, the priced frozen plan and execution-approval token in the same response. It does not run the study. Never ask for a separate configuration approval before pricing. For recommended dataset-backed cohorts, submit the catalog-mode panel_configuration returned by pulselake_preview_persona_cohort; individual panel.personas are neither required nor expected. Explicit persona records are required only for constructed or user-defined panels.
pulselake_copy_simulation
ChatGPTRead a previous study's editable configuration without copying execution approval. Uses only the authenticated user's own workspace; does not run a study or spend credits.
pulselake_explain_simulation_option
ChatGPTAnswer detailed user questions about how Persona Research, World Simulation, or a specific supported study works. Returns when to choose it, when not to, the approval sequence, controlled inputs, evidence and outputs, limitations, runtime and credit drivers, typical questions, and comparison guidance. Call this whenever a user asks how, why, what happens, what is produced, what it costs, or which option to choose. It is read-only and uses no credits.
pulselake_explore_persona_dimensions
ChatGPTBrowse the canonical PulseLake persona schema by category or search term. Returns stable dimension IDs, labels, categories, allowed values, and usage policy. Restricted dimensions are hidden unless explicitly requested. No credits are used.
pulselake_get_neural_response_quota
ChatGPTShow this authenticated user's remaining free Neural Response runs today and UTC reset time. All plans have two daily runs; this never charges credits.
pulselake_get_simulation
ChatGPTRead the authenticated user's tenant-scoped simulation job status, step progress, ETA, retry interval, hard deadline, and next action. This does not create another run or spend credits. Surface progress to the user between polls and stop polling immediately when terminal is true.
pulselake_get_simulation_artifact
ChatGPTRetrieve one named report artifact from a completed tenant/user-scoped simulation. Normally returns a private download_url for the file that expires after expires_in_seconds (request the artifact again for a fresh link). Only when no link is available is the file returned inline: then the first text block is the file itself (base64 for binary files); save it unchanged and never save the surrounding JSON response as HTML.
pulselake_get_simulation_balance
ChatGPTRead available, reserved and consumed credits. Uses only the authenticated user's own workspace; does not run a study or spend credits.
pulselake_get_simulation_result
ChatGPTReturn the completed study's structured public findings, settled job status, and available report artifact names. Fails safely while the job is incomplete.
pulselake_list_simulations
ChatGPTList the signed-in user's own past and current Simulation Research studies, newest first: job_id, study type, title, status, created and completed times, a short headline answer for completed studies, and available report artifact names. Pass next_cursor to page through older studies. Read-only; no credits are used. Use pulselake_get_simulation_result for a completed study's findings and pulselake_check_simulation_state for the live status of a queued or running study.
pulselake_open_simulation_studio
ChatGPTOpen the general PulseLake Simulations sign-in for studies, history and account management. Uses only the authenticated user's own workspace; does not run a study or spend credits.
pulselake_preview_persona_cohort
ChatGPTValidate canonical dimension names and values, calculate exact catalogue matches, inspect stratification cells, and surface feasibility, balance, restricted-dimension policy, and warnings before the user approves a panel. Its input schema describes only the requested cohort design; its result also returns panel_configuration, a complete catalog-mode object that pulselake_continue_simulation accepts unchanged. Dataset-backed approval covers the exact filters, strata, deterministic seed, and sample size—not individual records. Never fabricate panel.personas or convert this result to explicit_personas. No study credits are used.
pulselake_preview_simulation
ChatGPTValidate the complete study contract and return the exact branded execution plan, deliverables, aggregate credit estimate, and short-lived approval token. Does not execute work or reserve credits.
pulselake_preview_simulation_follow_up
ChatGPTPreview and price one evidence-grounded follow-up question about a completed multi-agent scenario simulation. The conversation handle is tenant/user scoped and no work runs until the user approves the estimate.
pulselake_read_simulation_report
ChatGPTRead an existing HTML report for the sandboxed viewer. Uses only the authenticated user's own workspace; does not run a study or spend credits.
pulselake_recommend_persona_dimensions
ChatGPTRecommend task-relevant canonical persona dimensions for a specific Simulation Research study and user-supplied context. Returns rationales and allowed values; the user must review the proposed design before cohort preview. No credits are used.
pulselake_recommend_simulation_study
ChatGPTRecommend Persona Research or World Simulation and the best matching study for the user's exact decision using deterministic outcome matching. Returns a recommendation, alternatives, and a focused clarification question when the request is ambiguous. Call this when the user describes a goal without naming a study. No credits are used.
pulselake_resume_simulation
ChatGPTFirst preview the credit limit for resuming a failed simulation. Show it and obtain explicit user approval, then call again with the returned approval_token and approved=true. Preserves completed stages.
pulselake_run_simulation
ChatGPTSubmit exactly the previewed tenant-scoped study after the user explicitly approves its total credit estimate. Reserves credits, returns a job ID, and never accepts an unpreviewed or modified study through the approval token. After submission, immediately follow next_tool until the job reports a terminal completed or failed status; never submit a replacement run.
pulselake_simulation_explore
ChatGPTExplain the two Simulation Research methods—Persona Research for independent responses and World Simulation for interacting actors—plus the complete catalog of supported studies, required configuration, outputs, workflow, and limitations. Call this before asking the user to choose a study. No credits are used.
pulselake_start_neural_response
ChatGPTStart the guided Neural Response video-research flow. Asks for video, audience profiles, profile confirmation and research-use acknowledgment, then previews zero credits and the two-runs-per-day limit. Does not start inference until explicit approval. Not a persona survey or world simulation.
pulselake_start_persona_research
ChatGPTStart the conversational Persona Research guide after the user confirms this method. Requires only the exact objective; if the research outcome has not been explicitly chosen, PulseLake explains and asks about each supported outcome. It returns the first typed configuration question and does not execute work, reserve credits, or require paid approval.
pulselake_start_simulation
ChatGPTLegacy compatibility entry point for clients already using the combined simulation_method and study_type contract. For new conversations, use pulselake_start_persona_research or pulselake_start_world_simulation. This starts configuration only and does not execute work or consume credits.
pulselake_start_world_simulation
ChatGPTStart the conversational World Simulation guide after the user confirms that interacting actors and an evolving scenario are required. Requires only the exact objective; the method and legacy study type are selected automatically. It returns one configuration question and does not execute work, reserve credits, or require paid approval.
