Stop Making Parts by Hand: The Flux AI Engineer Builds and Maintains Your Electrical and Mechanical Part Library

Ask the electrical librarian and the mechanical librarian how big the company part library is. You get the same two answers from both of them. The first answer is a number: 4,000 parts, or 12,000, or 30,000. The second answer is a story about where the parts are.

The electrical story: some parts are in an Altium integrated library that the first librarian made in 2011. Some are in OrCAD .olb files from the team that the company acquired. The power group uses PADS. The new product team uses KiCad. A contractor sent a folder of Eagle libraries. And three engineers keep their own copies of the same op amp, each with a different footprint.

The mechanical story is the same. Most parts are in a SolidWorks vault. The acquired team used Creo, and the tooling group uses NX. Supplier STEP downloads are in a shared drive, with file names such as model_final_v3.step. The same M3 x 8 socket head screw is in the library four times, with four part numbers. The display, the motors, the battery pack, and the harnesses are in a spreadsheet, with the supplier, the material, and the finish in columns that nobody checks.

Each of these parts was verified once. Somebody read the datasheet or the drawing, checked the pinout or the hole pattern, and built a product that worked. That work is valuable. But a library that is spread across many tools is slow to search, hard to keep correct, and expensive to maintain. Both teams wrote custom scripts to move data between the CAD tools, the spreadsheets, and the ERP system. When a distributor runs out of a part, or a supplier changes a drawing, a person finds out late and fixes the data by hand. Questions to suppliers go out as screenshots in email.

And somebody has to make every new part by hand. An electrical engineer reads a 100-page datasheet, draws the symbol, draws the footprint pad by pad, and checks it all again. A mechanical engineer measures a motor or a display from a drawing and models the STEP file in CAD. One part takes hours. A new product needs hundreds of new parts, so this work takes weeks to months. Most teams have no librarian, so design engineers do it between design tasks, and the product waits.

With Flux, electrical and mechanical library teams get an AI-first platform for all of this, for the first time. Board parts and mechanical parts are in one library. Every purchased part carries real-time inventory, lead times, and cost from distributors. Roles and permissions control who can change the library. Suppliers work with you in the same project, not in an email thread.

And in Flux, the AI engineer does the work for you. Give it an MPN, a datasheet, or the manufacturer's product page, and it makes the complete part: the symbol, the footprint, and the 3D model. When there is no STEP file, it makes the STEP model itself, from the drawing or from photos. It imports whole libraries at once and finds the duplicates. It fills in missing data, and it checks every part against its source. Nobody on the team makes parts by hand. The librarians stop copying data and start making decisions.

Import whole libraries, not one part at a time

Upload a folder, a ZIP, or a set of library files, and ask the AI engineer to import them. It reads every file, makes a list of the parts that it found, and tells you what each part has: a symbol, a footprint, and a 3D model. Then it imports the parts that you select, or all of them. Each part becomes a normal, editable Flux part in your account.

Flux reads the part libraries of every major EDA vendor:

• Altium: .SchLib, .PcbLib, and .IntLib
• Cadence OrCAD and Allegro: .olb symbol libraries and .dra footprints
• Siemens PADS: .asc, .d, .p, and .pcb
• KiCad: .lib, .kicad_sym, and .kicad_mod
• Autodesk Eagle: .lbr
• Autodesk Fusion Electronics: .flbr
• EasyEDA: Standard .json parts
• JEP30: .xml part data

Mechanical parts come in as STEP models, from SolidWorks, CATIA, Creo, NX, Onshape, Fusion, or a supplier website. The AI engineer makes a part for each model, with its MPN, its supplier, and its properties. The geometry stays editable: ask for a deeper flange, a new hole pattern, or vent slots, and the AI engineer edits the STEP solid.

You do not need library files at all. Give the AI engineer only an MPN, a datasheet, or a link to the product page on the manufacturer's website. It makes the complete part: the symbol, the footprint, and the 3D model for a board part, and the 3D model for a mechanical part. When the supplier has no STEP file, the AI engineer makes the STEP model itself, from the datasheet drawing or from photos. The section The AI engineer makes new parts and STEP models for you shows how.

You can also get parts from a design. Give the AI engineer an old Altium or KiCad project, and it creates one library part for each part in the schematic. When you have only a board file, it creates one footprint part for each footprint on the board. A shipped product becomes a library of parts that you know work.

Example prompts:

• "Import every part from these three Altium integrated libraries."
• "Here is a ZIP of our OrCAD symbol libraries and Allegro footprints. Show me what is in it, then import the power and the connector parts."
• "Here is a STEP export of the purchased parts in our SolidWorks vault. Make a part for each model, and take the MPN and the supplier from this spreadsheet."
• "Import these 40 supplier STEP files. Name each part by its MPN, not by its file name."
• "Here are 25 MPNs from our new BOM. Make each part from the manufacturer's product page, with its symbol, footprint, and 3D model."
• "This gear motor has no STEP file. Make the STEP model from the drawing in its datasheet."
• "Make library parts from every part in this KiCad project."
• "This PADS library has no 3D models. Here are the STEP files. Import the parts with their 3D models."

Import without duplicates

The first problem in every library migration is duplicates. Two teams made the same part. One part has a new name in each tool. When you import everything, you get everything twice. Mechanical libraries have this problem at scale: the same screw, washer, or standoff can exist under many part numbers.

The AI engineer checks the library before it imports a part. It searches by part name and by manufacturer part number. When the part is already in your library, it gives you the link to the existing part and does not import a second copy. You get one list of new parts and one list of parts that you already have.

• "Import this library, but skip every part that we already have."
• "Which of these 600 parts are already in our library? Give me the links."
• "We have three versions of the OPA2333. Compare their pinouts and footprints, and tell me which one matches the datasheet."
• "Find every M3 socket head screw in our library. Group them by length and standard, and list the duplicates."

The AI engineer makes new parts and STEP models for you

Every library has a gap that no import can fill: the part that is not in any library yet. It is a new power management IC that was released last quarter, a radio module from a supplier in another country, a gear motor that the mechanical team picked this morning, or a display with a custom bezel. Before Flux, an engineer stopped design work for half a day to read the datasheet or the drawing and model the part by hand. A new product needs hundreds of these parts, so this work took weeks to months.

In Flux, nobody makes these parts by hand. Give the AI engineer the MPN, the datasheet, the drawing, or the manufacturer's product page, and it makes the complete part, including the STEP model.

For a board part, it confirms the exact MPN, including its suffix, in the ordering table. Automotive, temperature-grade, and package variants often have different datasheets, so this step matters. It makes the symbol, with each pin named for its function, such as VIN, EN, or PGOOD, and with pin descriptions from the datasheet. It makes the footprint from the land pattern: from the package dimensions for SOIC, TSSOP, or QFN, and pad by pad for irregular lands, with exposed pads, shield pads, and oval drills. When the datasheet shows the land pattern in bottom view, it mirrors the drawing so that pin 1 is at the correct corner.

For a mechanical part, it reads the mechanical drawing. It builds a dimension table with a citation for each value: the outline, the datum, the hole pattern, the shaft, the flange, and the mating faces. Then it builds the 3D model from that table. Fasteners and inserts follow the standard for their size and head style. Flux is kernel agnostic, so the AI engineer builds each part on the CAD kernel that suits it: exact solids for machined parts, and smooth surfaces for hand-held products.

When a supplier has no CAD model and no drawing, the AI engineer can build the model from photos of the part with one known dimension. Flux publishes the first version of each new part, and the part is ready to use in the board, the enclosure, or the assembly.

Example prompts:

• "Make a part for the TPS62913RPUR from its datasheet, and place it in this design."
• "Here is the datasheet PDF of our new sensor module. Make the part. The land pattern is in bottom view."
• "Make a part for this 24 V gear motor from the drawing on page 3, with the shaft, the flange, and the four M3 mounting holes."
• "Make a part for this 5-inch touch display from its datasheet, with the bezel, the mounting holes, and the FPC connector pinout."
• "Make parts for M3 x 8, M3 x 12, and M3 x 16 stainless socket head screws, and a heat-set insert for each."
• "This connector has no STEP model. Make a 3D model from the package drawing on page 9 of the datasheet."
• "Here are photos of the sample relay from our supplier. Its body is 15.7 mm long. Make a 3D model of it."
• "These 12 MPNs from the new BOM are not in our library. Make a part for each one."

Enrich old parts with source data

Old libraries are full of gaps. A resistor has a value but no manufacturer part number. A regulator has pins named 1 to 8, so every engineer must open the datasheet to find the enable pin. A bracket has a model but no material and no finish. A bearing has a file name but no supplier. These parts work, but they are slow to use and easy to use wrong.

The AI engineer reads the datasheet or the supplier page of each part and fills in the gaps. For board parts, it adds the manufacturer, the MPN, the datasheet link, the package, the pin names and pin descriptions, and the key ratings, such as voltage, current, and temperature range. For mechanical parts, it adds the manufacturer, the MPN, the material, the finish, the thread, the standard, and the key ratings, such as torque, load, or IP rating. It cites the source for each value that it adds, so a reviewer can open the source page and check it.

• "Import these Eagle libraries from our contractor, and add the manufacturer and the MPN to each part from this spreadsheet."
• "This regulator has pins named 1 to 8. Rename each pin to its datasheet name, and add the pin descriptions."
• "In this project, add the manufacturer and the MPN to every resistor and capacitor that has only a value."
• "Add the material, the finish, and the thread standard to every fastener in this assembly."
• "Add the rated torque, the rated speed, and the shaft diameter to each of these 6 motors, with the datasheet page for each value."
• "List every part in this design that has no datasheet link, and find the datasheet for each one."

Verify every part against its source

A wrong part is expensive. A swapped pin on a regulator, a pad pitch that is 0.1 mm off, or a wrong package suffix on the purchase order can cost a board spin. A hole pattern that is mirrored, a shaft that is 1 mm too short, or a connector body that hits the enclosure wall can cost a tooling change. Most teams find these errors late: at assembly, at bring-up, or when the first parts come back from the supplier.

The AI engineer checks parts against the datasheet or the drawing before the design goes out. It checks each new part that it makes, and it checks any part that you point it to.

For board parts:

• Pins: the pin count, the pin names, and the pin types match the datasheet.
• Pads: each pin has a pad, and no pads overlap.
• Pad geometry: the pad size, the pitch, and the row span match the land pattern, and pin 1 is at the correct corner.
• Package: the package type and the body size match the datasheet.
• MPN: the part number and its variant suffix match the part that you will order.

For mechanical parts:

• Outline: the overall size and the datum match the drawing.
• Mating features: the hole pattern, the shaft, the flange, and the connector openings are at the drawing positions, with the correct orientation.
• Fit: in the assembly, the AI engineer measures the distance between parts, finds overlaps, and compares each number with your clearance rule.
• MPN: the part number, the size, and the material match the part that you will order.

When it finds a mismatch, it shows the source value and the part value, with the citation. It can fix the part, publish a new version, and update the designs that use it.

• "For U4, verify the pin count and functions, the package, the pad sizes, the pitch, the exposed pad, and the MPN variant against the datasheet."
• "We imported 300 parts from the old Altium library. Check the pinout of each IC against its datasheet."
• "Check the mounting hole pattern of this motor model against the supplier drawing."
• "Check that the battery pack has 2 mm clearance on every side in this enclosure."
• "The QFN footprint of U7 has the wrong exposed pad size. Fix it in the part, publish a new version, and update this design."
• "Does the MPN on J2 match the footprint and the 3D model? Check the ordering table in the datasheet."

Keep thousands of parts current

A library is not finished when the import is done. Parts go end of life. Distributors run out of stock. A manufacturer changes a package. A supplier changes a bracket drawing. The librarian finds a better footprint or a cleaner model and must update every design that uses the old one.

Flux connects every purchased part to real-time inventory, lead times, and cost from distributors. The AI engineer uses that data when it chooses parts, and it leaves end-of-life and obsolete parts out of its search results. When it finds a part that cannot be bought, it can find an in-stock part with the same fit and function and put it into the design.

When you publish a new version of a part, Flux tracks the version and what changed. Every project that uses the part shows that an update is available. The engineer reviews the update and applies it, one part at a time or all at once.

• "Find every part in this design that is end of life or out of stock, and suggest an in-stock replacement with the same footprint."
• "Which parts in this BOM have a lead time over 12 weeks? Suggest an alternative for each one."
• "Replace the obsolete LM317 parts in this design with an in-stock regulator, and keep the connections."
• "Our supplier moved the mounting holes on this bracket by 2 mm. Update the model and publish a new version."
• "Which parts in this BOM have only one source? Find a second source for each one."

Make the library follow your standards

Every professional team has library rules. Electrical teams have rules for designator prefixes, property names, symbol styles, and footprint naming. Mechanical teams have rules for part numbering, units, fastener standards, materials, and the data that each purchased part must have. The rules are usually in a document that nobody reads.

In Flux, you can give the rules to the AI engineer. It applies them when it imports, makes, and updates parts. A part that a new engineer imports on Friday looks the same as a part that the librarian made five years ago.

• "Our property names are Manufacturer, MPN, Datasheet, and Package. Use them for every part that you import."
• "Use the designator prefix U for every IC and J for every connector in this library."
• "Every fastener must have a standard, a material, a finish, and a length in millimeters. List the fasteners that do not."
• "Review this new part against our library rules, and list every difference."

Permissions, reviews, and suppliers in one place

A library is shared infrastructure. A wrong change to one footprint or one bracket model can reach every product that uses it. So library teams need control: who can change a part, who reviews the change, and when projects get it.

In Flux, organization roles and project-level permissions control who can view, comment on, and edit each part. A change to a part goes into a new published version. Projects keep the version that they use until an engineer reviews the update and applies it. Comments and @mentions stay on the part, so the review history stays with the data.

Suppliers join the same work. Invite a connector vendor, a machine shop, a molder, or a contract manufacturer to one project by email, with view, comment, or edit access. They see that project, not the rest of your organization. The connector vendor comments on the footprint itself. The machine shop marks the hole that it cannot drill at that depth. The contract manufacturer marks the pad that is too small for its process. Nobody sends screenshots over email, and nobody copies the answers back into a spreadsheet.

• "Summarize the open comments from our connector vendor on this part, and make the changes that they asked for."
• "The machine shop asks for a 1 mm radius on all inside corners of this bracket. Make the change and publish a new version."
• "List every part that changed since the last published version, and describe each change for the review."

Large jobs run in parallel

A library migration has thousands of small tasks. In Flux, the AI engineer can split a large job into parallel tasks: one task per library file, per part family, per vault folder, or per design. You see the progress of each task, and you get one report at the end: what was imported, what already existed, and what needs a decision from a person.

The librarians stay in control. The AI engineer does the work that is the same every time. The librarians make the calls that need experience: which of two footprints to keep, which screw standard to use, which second source to approve, and which old part to retire.

Example: one library from many tools

Here is how a team can move its electrical and mechanical libraries into Flux:

1. Collect. Put every library file into one folder: the Altium integrated libraries, the OrCAD and Allegro files, the PADS and KiCad libraries, the STEP exports of the purchased parts from the CAD vaults, and the purchased-parts spreadsheet. Add the old projects that shipped. Upload the folder as a ZIP.
2. Index. Ask: "List every part in these files, with what it has, and tell me which ones we already have in our library."
3. Import. Ask: "Import every part that we do not have yet. Use our property names, and take the MPN and the supplier of each mechanical part from the spreadsheet."
4. Fill the gaps. Ask: "Make a part from the datasheet or the drawing for each part in our BOMs that is not in any of these files."
5. Enrich. Ask: "For each new part without an MPN, a datasheet link, or a material, find it and add it."
6. Verify. Ask: "Check the pins and the pad geometry of every IC, and the mounting features of every motor and display, against the source. List the differences."
7. Clean up. Ask: "Find every part that is end of life, out of stock, or a duplicate, and suggest what to keep."

The result is one library in one place. Every electrical and mechanical engineer on the team searches the same parts, with the same data, and with real-time inventory, lead times, and cost. And nobody on the team spent weeks making parts and STEP files by hand. The AI engineer did that work.

Questions

Which part library formats can Flux import?

Flux imports part libraries from Altium, Cadence OrCAD and Allegro, Siemens PADS, KiCad, Autodesk Eagle, Autodesk Fusion Electronics, and EasyEDA, and JEP30 part data. Mechanical parts come in as STEP models from SolidWorks, CATIA, Creo, NX, Onshape, Fusion, and supplier websites. The AI engineer can also make library parts from the parts in a design file.

Does the Flux library hold mechanical parts?

Yes. Displays, motors, batteries, cables, fasteners, brackets, and enclosure parts are in the same library as the board parts, with an MPN, supplier data, and a 3D model. The AI engineer can make, edit, and check their 3D models.

Can Flux create a part from a datasheet or a drawing?

Yes. Give the AI engineer the MPN, the datasheet, or the drawing. For a board part, it makes the symbol, the footprint, and the 3D model. For a mechanical part, it makes the 3D model from a dimension table with a citation for each value. It can also build a model from photos. It checks the part and publishes it to your library.

How does Flux verify a part?

The AI engineer compares the part with its datasheet or drawing. For board parts, it checks the pins, the pads, the pad geometry, the package, and the MPN. For mechanical parts, it checks the outline, the mating features, the fit in the assembly, and the MPN. It cites the source for each value, and it lists each mismatch.

We manage our library in spreadsheets and scripts. Can Flux replace that?

Flux takes over the part work that those files do. The AI engineer reads your spreadsheets and makes or updates the parts. For every purchased part, Flux keeps real-time inventory, lead times, and cost, and it keeps versions and permissions with each part, in one library that the whole team uses.

Can I edit imported parts?

Yes. Each imported part is a normal Flux part. You and the AI engineer can edit its symbol, its footprint, its pins, its 3D model, and its properties, and you can publish new versions.

Does Flux import parts that are already in my library again?

The AI engineer checks your library first. When a part is already there, it gives you the link to the existing part and does not import a copy.

Can I export my library from Flux?

Yes. Flux exports part data as JEP30, an open standard for part data exchange, and mechanical models as STEP, STL, and 3MF. Your parts are not locked in.

How do I start?

Pick one real library that your team uses every day, electrical or mechanical. Upload it, and ask the AI engineer for an index and a duplicate check. Then compare the result with what your librarian expects. Talk to us about Flux for teams to plan a migration of your full library.