[BUG] gopher montecarlo with gopher key and 123180 on credit balance

[rahaar-alvarez]

gopher v0.2.16 CLI
2026/01/28 21:21:01 [Storage] Storage initialized at C:\Users\LenovoX1Carbon\Documents\Gopher\gopher.db
╔═══════════════════════════════════════════════════════════╗
║ Monte Carlo Validation ║
╚═══════════════════════════════════════════════════════════╝
Asset: TAO
Period: 2026-01-01 to 2026-01-25
Intervals: [15m 1h 4h]
Runs: 100

Running simulations...
📊 Loading data with 20-candle warmup (80.0 hours before start)
📊 Loaded 165 candles for hyperliquid:TAO (4h)
📊 Loaded 2608 candles for hyperliquid:TAO (15m)
📊 Loaded 657 candles for hyperliquid:TAO (1h)
📊 Loaded 2288 candles for hyperliquid:TAO (15m) from 2026-01-01 to 2026-01-25
📈 Loaded 6838 BBO ticks for hyperliquid:TAO (sampled @ 5m) from 2026-01-01 to 2026-01-25 [cloud]
[Checkpoint] Manager not configured

←[1m←[94m╔══════════════════════════════════════════════════╗←[0m
←[1m←[94m║ 🚀 Backtest: hyperliquid:TAO.P ║←[0m
←[1m←[94m╚══════════════════════════════════════════════════╝←[0m

←[2m21:21:59←[0m ←[32m[INFO ]←[0m [TAO] Period: 2026-01-01 → 2026-01-25
←[2m21:21:59←[0m ←[32m[INFO ]←[0m [TAO] Timeframes: 15m, 1h, 4h
←[2m21:21:59←[0m ←[32m[INFO ]←[0m [TAO] Eval Interval: 15m (2305 evaluations)
←[2m21:21:59←[0m ←[32m[INFO ]←[0m [TAO] Model:
←[2m21:21:59←[0m ←[32m[INFO ]←[0m [TAO] Balance: $10000.00 (10x leverage)
←[2m21:21:59←[0m ←[32m[INFO ]←[0m [TAO] Risk: 2.0% equity, 0.5% min stop, 2.0:1 target R/R, 1.0:1 min R/R
←[2m21:21:59←[0m ←[32m[INFO ]←[0m [TAO] BBO Mode: enabled (6838 ticks, sampled @ 1s) | Avg Spread: 2.18 bps | Slippage: 5.0 bps

time=2026-01-28T21:21:59.211-07:00 level=DEBUG msg="[toolSelection] Starting tool selection" tools_count=3 forceReasoning=false source.file=EfdOKIJ7eaT/tools.go source.L=972
time=2026-01-28T21:21:59.212-07:00 level=DEBUG msg="[pickTool] Starting tool selection" forceReasoning=false source.file=EfdOKIJ7eaT/tools.go source.L=434
time=2026-01-28T21:21:59.214-07:00 level=DEBUG msg="[pickTool] Using direct tool selection" source.file=EfdOKIJ7eaT/tools.go source.L=438
time=2026-01-28T21:21:59.447-07:00 level=WARN msg="Attempt to make a decision failed" attempt=1 error="error, status code: 401, status: 401 Unauthorized, message: You didn't provide an API key. You need to provide your API key in an Authorization header using Bearer auth (i.e. Authorization: Bearer YOUR_KEY), or as the password field (with blank username) if you're accessing the API from your browser and are prompted for a username and password. You can obtain an API key from https://platform.openai.com/account/api-keys." source.file=EfdOKIJ7eaT/tools.go source.L=290
time=2026-01-28T21:21:59.530-07:00 level=WARN msg="Attempt to make a decision failed" attempt=2 error="error, status code: 401, status: 401 Unauthorized, message: You didn't provide an API key. You need to provide your API key in an Authorization header using Bearer auth (i.e. Authorization: Bearer YOUR_KEY), or as the password field (with blank username) if you're accessing the API from your browser and are prompted for a username and password. You can obtain an API key from https://platform.openai.com/account/api-keys." source.file=EfdOKIJ7eaT/tools.go source.L=290
time=2026-01-28T21:21:59.615-07:00 level=WARN msg="Attempt to make a decision failed" attempt=3 error="error, status code: 401, status: 401 Unauthorized, message: You didn't provide an API key. You need to provide your API key in an Authorization header using Bearer auth (i.e. Authorization: Bearer YOUR_KEY), or as the password field (with blank username) if you're accessing the API from your browser and are prompted for a username and password. You can obtain an API key from https://platform.openai.com/account/api-keys." source.file=EfdOKIJ7eaT/tools.go source.L=290
←[2m21:21:59←[0m ←[31m[ERROR]←[0m [TAO] [2026-01-01 00:00] Evaluation error: LLM decision failed: LLM execution failed: failed to select tool: failed to pick tool: tool selection failed: failed to make a decision after 3 attempts: error, status code: 401, status: 401 Unauthorized, message: You didn't provide an API key. You need to provide your API key in an Authorization header using Bearer auth (i.e. Authorization: Bearer YOUR_KEY), or as the password field (with blank username) if you're accessing the API from your browser and are prompted for a username and password. You can obtain an API key from https://platform.openai.com/account/api-keys.

[teslashibe]

@rahaar-alvarez thanks for posting this issue - workin on this

[teslashibe]

Investigation Update

I've done a forensic code review and found some important clues:

Key Finding

The log shows [TAO] Model: with an empty value - this indicates the engine config's Model field isn't being set properly.

Root Cause Hypothesis

There appears to be a disconnect between the resolved inference configuration and the backtest engine config. When running Monte Carlo:

1. Gopher key is detected ✓
2. Inference resolves correctly ✓
3. BUT the EngineConfig.Model field remains empty
4. This causes the go-openai library to default to api.openai.com instead of Gopher API

Questions to Help Reproduce

1. What's in your config file? Could you share the contents of:

   • Windows: C:\Users\LenovoX1Carbon\Documents\Gopher\config.json
   • (Remove any sensitive API keys before sharing)

2. Full CLI command: What exact command did you run? For example:

   gopher montecarlo --asset TAO --start 2026-01-01 --end 2026-01-25

3. First run or existing config?

   • Did you run the desktop app first to configure settings?
   • Or is this a fresh install with only BART_GOPHER_CODE set?

4. Environment setup: How did you set your Gopher key?

   • Via environment variable: export BART_GOPHER_CODE='gopher_xxx'
   • Or another method?

This will help us reproduce and fix the issue. Thanks!

[teslashibe]

Root Cause Identified

After forensic code review, we found the root cause:

The Bug

The Resolved.Validate() function that checks for empty ModelID was added locally but was NOT included in the v0.2.16 release.

Without this validation:

1. Empty BacktestModelID from config passes through silently
2. EngineConfig.Model is set to empty string (explains [TAO] Model: with no value in your logs)
3. The go-openai library defaults to api.openai.com when model/baseURL are empty
4. Your Gopher key is sent to OpenAI's direct API, causing the 401 Unauthorized

The Fix

The validation code has already been written locally and needs to be released:

• Resolved.Validate() with ModelID check
• IsValidLoop() and IsValidProvider() functions
• Defensive validation in NewLLM() constructor

Workaround (until next release)

Run the desktop app first and configure your settings there. This ensures the config file has proper model IDs set before using the CLI.

Next Steps

• We'll include the validation in the next release
• The fix will provide a clear error message if model ID is missing instead of silently failing

Thanks for reporting this - it helped us identify a gap in our validation layer!