How StreetTunedAI works
StreetTunedAI Portal: How the Review Process Works
This is not a sales post. It is a straight explanation of what happens from the time you save a scanner log to the time you read the review.
A lot of people hear “AI tuning assistant” and assume it means it is going to tune their truck, write a file, or tell them to flash something without understanding the vehicle. That is not what this is.
StreetTunedAI is a tuning assistant built to be a second set of eyes on your real scanner data. It looks at the channels in the log, checks whether the log is good enough to answer the question, flags safety issues first, and gives you a structured review of what the data supports, what is missing, and what should be left alone until there is better evidence.
You still make the final calibration change. You still decide when the vehicle is ready for a road pull. You still own the tune. The portal is there to keep you from guessing, chasing the wrong thing, or making a change because one number looked strange for half a second.
Step 1: Start with the question you are trying to answer
Before you make a log, be honest about where you are in the tune. A good log starts with a clear purpose.
Examples:
•“I just need to see startup and idle.”
•“I am checking part-throttle fuel trims and MAF behavior.”
•“I need to verify whether this knock retard repeats in the same area.”
•“I am ready for a WOT pull and want to look at commanded versus actual fueling, pressure tracking, and KR.”
•“I changed something and want to compare the next log to the last one.”
Do not try to answer every tuning question in one drive. Startup, idle, steady cruise, transitions, and WOT are different parts of the job. A clean log for the specific tuning step is worth more than a giant random drive with every condition mixed together.
Step 2: Use the right scanner channels before you drive
The portal can only review the evidence that is actually in the CSV. If the needed channel was not logged, StreetTunedAI does not pretend it was there.
For a basic startup, idle, or part-throttle review, the useful channels normally include engine RPM, throttle or pedal position, MAP/load, MAF frequency where applicable, short-term and long-term fuel trims, spark, knock retard, coolant temperature, intake air temperature, closed-loop state, PE status, and DFCO status.
For a WOT review, you need the WOT evidence too: throttle/pedal, RPM, load or MAP, commanded fueling, actual wideband fueling from a known and correctly mapped controller, spark, knock retard, high-side actual and desired fuel pressure where applicable, and low-side fuel pressure if the vehicle logs it.
If you use an external analog wideband, make sure the controller and formula are known. A wideband conversion that is not confirmed—or a sensor signal that is flatlined, implausible, or not behaving like a real signal—does not get used for fuel correction direction.
That is not the portal being difficult. That is the point. Bad input should stop bad advice.
Step 3: Record the type of log needed for your current tuning step
If you are working on startup, idle, or cruise, do not jump straight into a WOT pull just because you can. Start with the part of the tune you are actually working on.
For steady part-throttle work, the goal is stable evidence. Let the vehicle warm up. Hold a reasonable steady condition where it is safe. Avoid using a hard acceleration, a lift, a shift event, DFCO, or active PE as though it were a clean fuel-trim sample.
For a WOT review, only do it when the vehicle is ready, the road is safe, and you have the right data being logged. A useful pull needs to be a real pull, not a quick stab at the throttle. The review looks for enough duration, RPM movement, valid WOT samples, usable MAF/load bins, pressure behavior, commanded versus actual fueling, and KR context.
One important point: two pulls are not automatically one pull just because they happened on the same drive. If you lift and get back in it, the assistant keeps them separate. If the same WOT evidence continues through a gear change, it stays one continuous pull. Brief missing or malformed throttle data can be treated as a data dropout only when the surrounding evidence supports that conclusion. A real throttle release is a real boundary.
Step 4: Export the CSV from your logging software
StreetTunedAI reviews a CSV datalog, not a screenshot and not a tune file.
Save or export the scanner log as a CSV from HP Tuners, EFILive, or Haltech. Keep the original file name if it helps you remember the drive. A useful file name might include the date and what you were testing, such as:
2026-08-19_idle_cruise_after_maf_change.csv
or
2026-08-19_wot_pull_93octane.csv
You do not need to clean up the file by hand. Upload the real log. The portal checks channel names, normalizes common scanner naming differences, and tells you what it found.
Step 5: Upload the real log to your StreetTunedAI review access
After you sign in, choose the vehicle if you have more than one. Then upload the CSV for the vehicle and the tuning step you are working on.
Selecting the right vehicle matters. A follow-up log makes more sense when it stays tied to the same vehicle and the same work. If you are working through a MAF issue on one truck and a startup issue on another, keep those separate.
Once the CSV is uploaded, the StreetTunedAI tuning assistant reads the file through its deterministic analyzer first. That analyzer checks the log structure, channel evidence, operating conditions, WOT segmentation, fuel-pressure tracking, wideband provenance, and safety gates. The AI explanation comes after that structured review. It is not supposed to invent a table, make up math, or override a failed safety gate.
Step 6: Checks come before conclusions
This is where the portal earns its keep.
Before it tells you what the data supports, it checks whether the log is good enough to support that answer. Depending on the log, that can include:
•Whether the required channels are present.
•Whether the log includes a usable time base.
•Whether the log is in closed loop, PE, DFCO, a transient condition, or a stable condition.
•Whether WOT actually happened.
•Whether a WOT event is long enough and covers enough RPM to be meaningful.
•Whether wideband provenance is confirmed.
•Whether the wideband signal itself looks believable.
•Whether high-side and low-side fuel-pressure data is present and tracking correctly.
•Whether knock retard is isolated, repeated after recovery, or sustained.
•Whether torque intervention, traction events, a shift, or low delivered spark could explain a KR event.
If the log does not have what it needs, the report should say that plainly. “Need a better log” is not a failure. It is better than getting a fake answer and changing the wrong thing.
Step 7: Read the review in order
The portal report is laid out so you can start with the important part instead of digging through a wall of raw scanner data.
What I Received
This tells you what the portal confirmed, what looks suspect, and what is missing. It is the inventory of the evidence. If you did not log wideband, pressure, trims, or knock retard, it should show up here instead of being quietly ignored.
What I See
This is the plain-language summary of the pattern in the log. It should tell you what was seen in the actual conditions that were captured, not what somebody hopes the vehicle is doing.
Fuel Trim Analysis
For stable closed-loop evidence, the review looks at both banks. It does not hide a big bank-to-bank difference by averaging it into a number that looks harmless. When the evidence is solid, it can show the raw trim direction, the first-pass bounded direction, and the MAF-frequency area involved. When the banks disagree too much, it should block the combined direction and tell you that the disagreement needs to be understood first.
WOT / Power Enrichment Review
The report keeps each WOT pull separate. You can see the pull number, time range, sample count, RPM movement, load range, fueling evidence, pressure evidence, KR evidence, and whether that pull was accepted or rejected.
WOT fuel direction is not issued just because there was a wideband column in the file. The controller/formula has to be confirmed, the signal has to be plausible, the pull has to be good enough, fuel pressure has to track, and the applicable pulls have to agree where they overlap. If the path is blocked, you get the block reason—not a number you are tempted to use anyway.
Knock / Spark Review
KR is not treated as all the same thing. A single isolated event is not the same as repeated events after recovery. A continuous run is not described as six separate knock events just because there are six rows in the log.
The review also looks for torque intervention, shift context, low spark, RPM/load location, and whether the KR repeats in the same area. Protective timing guidance, when that separately tested path is enabled in the future, is limited and clearly marked as protective. It does not prove the root cause. It does not add timing. It does not replace checking fuel, sensor, mechanical, or operating-condition causes.
First Edit Path
This tells you the safest supported next move for the log you uploaded. Sometimes that is a small, bounded data-supported direction. Sometimes it is “do not edit yet.” Sometimes the right answer is to add a missing channel and make another log.
Do Not Touch
This section matters. It tells you what should not be changed from that log. The tuning assistant is not supposed to send you after injector timing, torque modeling, dynamic airflow, high-pressure pump calibration, direct torque-model changes, or timing additions when the evidence does not support it.
Next Log Plan
This is the practical next step: what to log next, which conditions to capture, and what question the next file needs to answer.
What StreetTunedAI is
•A tuning assistant for reviewing datalogs.
•A scanner setup check.
•A safety check before conclusions.
•A second set of eyes on the log you actually made.
•A structured way to separate confirmed evidence, suspect evidence, missing evidence, and unsupported guesses.
•A tuning assistant built to help you make better decisions with your own calibration work.
What StreetTunedAI is not
•It does not tune the vehicle for you.
•It does not flash your ECM, PCM, or TCM.
•It does not edit your tune file.
•It does not replace a professional tuner or your own judgment.
•It does not turn a bad log into a trustworthy answer.
•It does not guess when the needed data is missing.
•It does not give you a number to use when the evidence gates say the path is blocked.
The short version
Log the right channels. Record the right condition. Export the real CSV. Upload it to the right vehicle. Let the assistant run the checks first. Read what the data supports. Leave alone what the data does not support. Then make the next log with a purpose.
That is the whole deal. StreetTunedAI is there as a tuning assistant to help you move through a project with better evidence and fewer blind gu