[Bug] Simulation stuck in "Running" state after completing max rounds (10/10), "Generate Report" button remains disabled

[ste22-ai]

The simulation finishes all rounds (Round 10/10 reached in both Plaza and Community modules), but the frontend remains stuck in the Running state and does not enable the "GENERATE REPORT" button.

The backend gets stuck in an infinite polling loop for run-status endpoints without updating the state to completed.

Environment:
OS: Windows / Linux

Python Version: 3.12

Node.js Version: 20+

LLM_API_KEY=ollama
LLM_MODEL_NAME=qwen2.5:7b

Browser: Google Chrome

Steps to Reproduce:
Start a simulation with maxRounds=10.

Wait for the execution to process through all rounds (R1/10 to R10/10).

Observe the Simulation Monitor logs and status indicators once round 10 finishes.

Expected Behavior:
Once R10/10 finishes processing, the simulation status should update to Completed (or stop running), and the GENERATE REPORT button should become clickable/enabled.

Actual Behavior:
The Simulation Monitor logs show that Round 10/10 has finished:

Plaintext
15:58:41.447 [Plaza] R10/10 | T:0h | A:7
15:58:41.447 [Community] R10/10 | T:0h | A:10
The status indicator at top right remains in Running state.

The GENERATE REPORT button remains grayed out / disabled.

The backend console loops endlessly receiving status checks:

Plaintext
[backend] 127.0.0.1 - - [08/Oct/2026 16:16:01] "GET /api/simulation/sim_254be34773e1/run-status HTTP/1.1" 200 -
[backend] 127.0.0.1 - - [08/Oct/2026 16:16:01] "GET /api/simulation/sim_254be34773e1/run-status/detail HTTP/1.1" 200 -
[backend] 127.0.0.1 - - [08/Oct/2026 16:16:03] "GET /api/simulation/sim_254be34773e1/run-status HTTP/1.1" 200 -

[nyanduong]

This looks like a lifecycle issue rather than a round-counter issue. In the code, twitter_completed and reddit_completed are per-platform signals; the runner publishes a terminal status only after the subprocess exits and any enabled graph-memory finalization finishes.

A similar post-round process-wait behavior was documented in #779. Could you share a sanitized response from GET /api/simulation/<simulation_id>/run-status, especially runner_status, twitter_completed / reddit_completed, process_pid, and error, plus the last 50–100 lines of simulation.log?

If both platforms are complete and the PID is still alive, that would be consistent with the post-round command-wait behavior described in #779. If the status is stopping, and graph-memory updates are enabled, Zep ingestion may still be draining. If the process has exited but the status remains non-terminal, that points to a different state-update/finalization problem.

I wouldn’t apply --no-wait blindly: the project also needs to preserve the OASIS environment for interview/survey work before report generation, as discussed in #763 and PR #799.

[ste22-ai]

Hi thank you for helping me.
Here is the JSON response from GET /api/simulation/sim_5779d4e24c59/run-status:

{
  "data": {
    "completed_at": null,
    "current_round": 10,
    "error": null,
    "process_pid": 17508,
    "progress_percent": 100.0,
    "reddit_actions_count": 2,
    "reddit_completed": true,
    "reddit_current_round": 10,
    "reddit_running": false,
    "reddit_simulated_hours": 0,
    "runner_status": "running",
    "simulated_hours": 0,
    "simulation_id": "sim_5779d4e24c59",
    "started_at": "2026-10-09T19:34:20.037145",
    "total_actions_count": 4,
    "total_rounds": 10,
    "total_simulation_hours": 72,
    "twitter_actions_count": 2,
    "twitter_completed": true,
    "twitter_current_round": 10,
    "twitter_running": false,
    "twitter_simulated_hours": 0,
    "updated_at": "2026-10-09T19:35:52.547630"
  },
  "success": true
}

Here the simulation.log :
[19:35:38] ============================================================
[19:35:38] OASIS Dual-Platform Parallel Simulation
[19:35:38] Configuration File: D:\Download\MiroFish-main (2)\MiroFish-main\backend\app\services\../../uploads/simulations\sim_5779d4e24c59\simulation_config.json
[19:35:38] Simulation ID: sim_5779d4e24c59
[19:35:38] Wait Command Mode: Enabled
[19:35:38] ============================================================
[19:35:38] Simulation Parameters:
[19:35:38]   - Total Simulation Duration: 72 hours
[19:35:38]   - Time per Round: 60 minutes
[19:35:38]   - Configured Total Rounds: 72
[19:35:38]   - Maximum Round Limit: 10
[19:35:38]   - Actual Executed Rounds: 10 (Truncated)
[19:35:38]   - Agent Count: 3
[19:35:38] Log Structure:
[19:35:38]   - Main Log: simulation.log
[19:35:38]   - Twitter Actions: twitter/actions.jsonl
[19:35:38]   - Reddit Actions: reddit/actions.jsonl
[19:35:38] ============================================================
[19:35:38] [Twitter] Initializing...
[19:35:45] [Reddit] Initializing...
[19:35:51] [Twitter] Environment Started
[19:35:51] [Reddit] Environment Started
[19:35:51] [Twitter] Published 2 initial posts
[19:35:51] [Twitter] Rounds Truncated: 72 -> 10 (max_rounds=10)
[19:35:51] [Twitter] Simulation loop completed! Time elapsed: 0.1s, Total actions: 2
[19:35:51] [Reddit] Published 2 initial posts
[19:35:51] [Reddit] Rounds Truncated: 72 -> 10 (max_rounds=10)
[19:35:51] [Reddit] Simulation loop completed! Time elapsed: 0.0s, Total actions: 2
[19:35:51] ============================================================
[19:35:51] Simulation loop completed! Total time elapsed: 13.0 seconds
[19:35:51] 
[19:35:51] ============================================================
[19:35:51] Entered wait command mode - Environment remains running
[19:35:51] Supported commands: interview, batch_interview, close_env
[19:35:51] ============================================================