GENERAL M / CONTROL CENTRE
Overview
COLONEL ONLINE
SPEND TODAY
$1.42
$7.81 last 7 days
ACTIVE BUILDS
3
2 executing now
NEEDS YOU
1
Credential request
OPERATION SUCCESS
92.4%
Last 50 operations
COST — LAST 7 DAYS
SEE USAGE →COLONEL ACTIVITY
LIVEACTIVE BUILDS
CLICK ANY BUILDGENERAL M / BUILDS
Builds
TOTAL BUILDS
—
ACTIVE
—
WAITING / PAUSED
—
COMPLETE
—
BUILD MODELObjective → Contract → Operations → Artifacts → Acceptance → Complete
Connecting to General M API…
GENERAL M / EVENTS
Activity
GENERAL M / ECONOMICS
Usage & Costs
TOTAL SPEND
$24.19
30 days
AVG / BUILD
$1.47
Completed builds
AVG / OPERATION
$0.064
All components
LLM CALLS
386
$18.72 model spend
SPEND — 30 DAYS
DAILYSPEND BY BUILD
30 DAYSAVERAGE COST BY COMPONENT
PER INVOCATIONCOMPONENT ECONOMICS
ALL BUILDS| COMPONENT | CALLS | AVG | TOTAL | % SPEND |
|---|---|---|---|---|
| B1 Control | 184 | $0.009 | $1.66 | 6.9% |
| B2 Retriever | 68 | $0.021 | $1.43 | 5.9% |
| B3 Planner | 41 | $0.074 | $3.03 | 12.5% |
| B4 Builder | 79 | $0.143 | $11.30 | 46.7% |
| B6 Tester | 96 | $0.032 | $3.07 | 12.7% |
| B7 Debugger | 38 | $0.097 | $3.69 | 15.3% |
MODEL SPEND
LLM ONLYREPAIR ECONOMICS
QUALITY SIGNALAVG REPAIRS / BUILD
1.8
REPAIR SPEND
$4.82
FIRST-PASS RATE
61%
This section is designed to reveal when Colonel is spending too much repeatedly repairing an approach instead of replanning.
GENERAL M / EXECUTION ENGINE
Colonel
ONLINE
CURRENT LOAD
2
Executing builds
QUEUE
1
OPS TODAY
37
FAILURES TODAY
3
COMPONENT STATUS
ARCHITECTURE| COMPONENT | STATE | ROLE |
|---|---|---|
| B1 Controller | BUILDING | Routing / checkpoints / intervention |
| B2 Retriever | READY | Context retrieval |
| B3 Planner | NEXT | Planning / replanning |
| B4 Builder | PENDING | Workflow construction |
| Tester | PENDING | Evidence / validation |
| Debugger | PENDING | Diagnosis / repair |
ASK COLONEL
INFORMATION ONLYThis chat does not change execution. Use a build's Instruct Colonel panel to intervene.
CREATE BUILD
B0 BUILD CONTRACTOne build owns one objective. This form mirrors the current B0 contract input; the contract can evolve later as B3 Planner and B4 Builder are finalised.
NAME
DESCRIPTION
OBJECTIVE
SUCCESS CRITERIA
CONSTRAINTS
ACCEPTANCE CRITERIA
PERMISSIONS
Budget limit (USD)
Intelligence policy
B0 payload: action · request_key · name · description · objective · budget_limit_usd · constraints[] · permissions{} · success_criteria[] · acceptance_criteria[]
CONTRACT READY
Build will be submitted to B1
Prototype only — the live version will create the build/contract records and wake Colonel.
GENERAL M / MODEL INTELLIGENCE
Model Performance
MODELS TRACKED
12
MODEL RUNS
184
AVG COST / RUN
$0.041
BEST PASS RATE
94%
MODEL PERFORMANCE
SOURCE · general_m.model_performance| MODEL | CLASS / ROLE | INTELLIGENCE | RUNS | PASS RATE | AVG COST | AVG LATENCY |
|---|---|---|---|---|---|---|
| Model A | Builder | High | 31 | 94% | $0.082 | 18.4s |
| Model B | Planner | Medium | 47 | 91% | $0.037 | 8.7s |
| Model C | Controller | Low | 106 | 88% | $0.009 | 2.1s |
COST VS PASS RATE
MODEL COMPARISONWHY THIS EXISTS
Colonel can use observed performance by task/class and intelligence level instead of assuming that the most expensive model is always the best choice.
Model → role/class → intelligence level → attempts → success/failure → cost → latency → quality evidence
GENERAL M / CONFIGURATION
Settings
NOTIFICATIONS
| Telegram alerts | Enabled |
| Human intervention alerts | Enabled |
| Build completion | Enabled |
| Routine operation events | Disabled |
DEFAULT BUILD POLICY
Default budget$0.50
Repair ceiling3 attempts
Budget exhaustionPause + request
Human requestPause build
DEFAULT INTELLIGENCE LEVELS
OVERRIDABLE PER BUILD| COMPONENT | DEFAULT | ESCALATION |
|---|---|---|
| B1 Controller | Medium on uncertainty | |
| B2 Retriever | None | |
| B3 Planner | High on replan | |
| B4 Builder | High after failed attempt | |
| Tester | Medium if ambiguous | |
| Debugger | High after repeat failure |
These are Colonel's system defaults. A build contract can override them, and escalation rules can raise intelligence when uncertainty or repeated failure warrants it.
CONNECTED RESOURCES
| Supabase | Connected | State / memory / telemetry |
| n8n | Connected | Execution environment |
| Telegram | Connected | Notifications |
| OpenRouter | Connected | LLM routing |
GENERAL M / BUILD
Overview
Live
Activity
Outputs
Costs
Contract