Skip to main content

Biocognia Intelligence

One AI workspace over the BioCognia products, opened with one command. Installing it installs the products, and installing a product installs it, so any one of these gives you the app, SCIMAP Pro and Plexora in one environment:

pip install biocognia-intelligence     # or scimappro, or plexora
biocognia

Each product is used where its licence is on this device. (pip install biocognia is the licensing library alone, not the app.)

biocognia opens the Workspace in your browser. Start with a question, with your data, or both: drop or choose a file or folder, or connect a data location, and Send makes a workspace for them. The product that reads that kind of data opens it, and the assistant helps you explore and interpret it. Your workspaces, and SCIMAP Pro's own sessions, are listed together. A product with an application of its own, such as Plexora, opens in its own window from File › Applications.

There is one kind of workspace, and it is the app's. The products you have installed are its tools: SCIMAP Pro's analyses run in it, and Plexora's tools join its conversations. You never choose which product a workspace is for.

The app needs no licence of its own. It works once one product is installed and this device is connected with that product's licence: Tools › Connect this device in the page, or biocognia connect, shows a code you approve in the portal, and every product you hold a seat for is connected at once. The Tools menu shows each product's state (not installed, not connected, connected, with or without AI and MCP access) and the tools it brings.

One conversation can use several products. With Plexora installed, the assistant can also gate markers against the staining, check the image's quality, draw regions and look at rendered images in Plexora, and bring the results back into the table. With a Plexora Pro licence it can also run Plexora's AI gating and AI QC: the whole run on Plexora's own AI (Plexora's credits), or step by step, deciding from each packet's images itself.

You have The app gives you
SCIMAP Pro SCIMAP Pro's analyses in conversation, paid by SCIMAP Pro's licence
SCIMAP Pro and Plexora both in one conversation, Plexora's Free tools included
both, with Plexora Pro Plexora's AI gating and QC too
Plexora only Plexora's tools in conversation, paid by Plexora's licence, and images you add made Plexora projects; biocognia install scimappro adds the analyses

Coding agents such as Claude Code reach SCIMAP Pro over MCP where its licence includes MCP access: scimappro ai init wires them to scimappro mcp serve, and Tools › Connect a coding agent lets one into an open workspace.

The assistant can also work on the computer itself, as you would at a terminal: make a conda environment or a Python venv, install a package, run a command or a script (in the background for a long one, which keeps running if the app closes), and read, search and write files. Its shell keeps its folder from one command to the next.

Each workspace has a folder of its own, made when the workspace is, under ~/Biocognia (Preferences › Workspace folders changes where; the composer's menu › Workspace Folder… moves one). results/ holds what the work produced -- figures, tables, AnnData objects, scripts; SCIMAP Pro's figures and tables go there too -- and system/ what can be made again: downloads, command logs. Environments are made where their tools keep them (conda's own envs folder, ~/.virtualenvs for a venv), so conda activate and your editor find them, and the workspace records each one its work made. Deleting a workspace asks whether to delete either part; its system files take those environments with them.

Routine work runs without asking. You are asked before anything that reaches beyond the work: deleting outside the workspace's folder, uninstalling from an environment it did not make, installing into the app's own Python, sudo, killing a process, running a downloaded script, reading a credential -- and, unless autonomy is automatic, changing a file of yours outside the workspace's folder. Manual autonomy asks before every change. BIOCOGNIA_COMPUTER=0 turns this off.

Command What it does
biocognia Opens the app. --no-browser prints its address instead, and --port serves elsewhere (default 8480).
biocognia tools Each product as the Tools menu shows it: installed, connected, AI and MCP access, its tools. --json prints it as JSON.
biocognia connect Connects this device: a code to approve in the portal, or a licence key (biocognia connect BIOC-...). The same as biocognia license activate.
biocognia list The products installed, and those of the family that are not. --licence adds whether each one's AI licence lets a conversation run and how many of a guest's tools it unlocks; --json prints it all as JSON.
biocognia open plexora Opens a product's own application, for a product that has one.
biocognia install plexora Installs a product into the Python the app runs in (pip, else uv). --upgrade takes it to its newest release. A desktop application's own Python is updated by its installer instead.
biocognia doctor Checks each product: its registration, whether it can pay for the assistant, MCP access, its tools and whether it is running. Exits 1 on a problem, with what to do.
biocognia license status This device's licences, for every product. activate, refresh, install and the other verbs work as in each product's own command line.
biocognia serve --desktop A desktop shell's sidecar: one ready line, then the app until stdin closes.

Desktop app

desktop/ is the app as a desktop application (Tauri 2), moved here from SCIMAP Pro, whose own app it was until 0.1.27. It runs biocognia serve --desktop from a Python runtime inside the installer and shows it in a native window, with a menu bar, one instance at a time, .h5ad files and biocognia:// links opened in it, and Check for Updates. The runtime carries the app and SCIMAP Pro (scimappro[ai,bridge,search], a compiled wheel), so SCIMAP Pro's analyses are there with nothing else to install. The installer updates that Python, never pip.

python tools/release.py doctor     # what the toolchain lacks: uv, Node, Rust, and Xcode, MSVC or WebKitGTK
python tools/release.py all        # tests, wheel, runtime, installer, validate -> build/desktop, release/
python tools/release.py --scimappro ../scimappro/build/release/dist/scimappro-<version>-cp313-<platform>.whl all

SCIMAP Pro comes from PyPI unless --scimappro names a wheel or another requirement (SCIMAP Pro's python tools/release.py wheel builds one). A v* tag builds the Windows, macOS (Apple silicon) and Linux installers in .github/workflows/desktop.yml and publishes them, with the updater's latest.json, as a GitHub release. After changing the version in pyproject.toml, python tools/release.py propagate copies it into the shell's files; the tests fail until it does.

Variable What it does
BIOCOGNIA_DESKTOP_RUNTIME The runtime folder the shell runs (<runtime>/bin/python3, or <runtime>\python.exe on Windows) in place of the bundled one.
BIOCOGNIA_DESKTOP_PYTHON, BIOCOGNIA_DESKTOP_CWD Run the server with this Python, from this folder (a checkout's src), to work on the shell against a checkout: cargo run --manifest-path desktop/src-tauri/Cargo.toml.
BIOCOGNIA_DESKTOP, SCIMAPPRO_DESKTOP Set by the shell for the server. biocognia install refuses to run pip inside the bundled runtime.
BIOCOGNIA_SCIMAPPRO The SCIMAP Pro that tools/release.py puts in the runtime, as --scimappro does.
TAURI_SIGNING_PRIVATE_KEY, BIOCOGNIA_UPDATER_PUBKEY The updater's key pair (a secret and a repository variable). Without them, installers are built but update by download, not in place.

How it is put together

  • A workspace is the app's. One made here is a folder under the app's own (workspaces/), with biocognia.json saying which products are attached and who pays. The product that runs in this process (SCIMAP Pro) keeps its session in that folder, exactly as in its own sessions. SCIMAP Pro's own sessions, such as those its MCP server makes, are listed beside them and opened in place.
  • One product pays. Every model call of a workspace goes through the payer's own gateway, as that product's task (scimappro.chat.turn). The payer is the first product, SCIMAP Pro before Plexora, whose AI licence is on this device when the workspace is made; the person can switch it in the Tools menu, and it never changes by itself. Work the payer's licence does not cover (plans, on Plexora's) is refused with a sentence saying who could pay for it. There is no Intelligence licence: AI comes with each product's licence.
  • Other products join as guests. A guest's tools are listed in the conversation as <product>.<tool>, from a manifest shipped with the product, so the conversation's catalogue does not change when the guest is not running. Each call goes over spatialbridge to the copy already running, or starts one. The guest checks the call against its own licence. A tool that needs a licence is offered only where the guest's licence on this device covers it. Images a guest draws reach the model as images. Guests offer reads and reversible writes only, because the bridge cannot carry an approval. The home's own hand-offs, such as SCIMAP Pro's view_image, still move data between products, and the guest tools they already drive are not offered a second time. A file a guest takes, such as an image for Plexora, is handed to it through one of its tools when you add it to a workspace: Plexora makes it a project, and the conversation is told which. In a workspace with no product in this process (Plexora only), the guests' tools are all there is.
  • This computer is a tool too. computer.<tool> runs shell commands (in the foreground, or in the background to follow later, even after a restart), reads, lists, searches, writes and edits files, and finds the Python environments here: this app's own, conda, mamba or micromamba and their environments, venvs, and uv. It runs in this process, so a call can wait for the person's approval as a provider's own do, and its class is read off what the call would do: routine work is a reversible_write, a file of the person's outside the workspace's folder a source_file_write, and what reaches beyond the work destructive. A command runs non-interactively in its conversation's shell folder, without this app's credentials, and with conda on its PATH even when the app was started from the desktop. The contract is in src/biocognia_intelligence/computer.py.
  • A workspace has a folder. Chosen when it is made, recorded in the workspace with the environments its commands made, shown and moved from the page, and offered part by part when it is deleted (src/biocognia_intelligence/folders.py). An in-process provider whose host has use_folder(path) is told where its results go.
  • The runtime is shared. The conversation, plans, approvals and the page are spatialagents. Licences and the AI token come from biocognia.
  • Products register themselves. A product joins by registering one object under the biocognia.products entry point, from a package small enough to import in a moment. The contract is in src/biocognia_intelligence/registry.py.
[project.entry-points."biocognia.products"]
myproduct = "myproduct_intelligence:PRODUCT"

Adding a product

A new package joins without a release of this app:

  1. Depend on biocognia-intelligence>=0.3.1,<0.4, so the app comes with it. A family product is also added to the app's own dependencies, so it comes with the app and with every other product.
  2. Register one object under biocognia.products (above), and assert registry.check_registration(PRODUCT) == [] in its suite.
  3. Run in this process as a provider, whose provider().attach(ctx) answers a workspace's parts (its host, its conversations' tools and prefix, its plans) made with the gateway the app hands it, or join over spatialbridge as a guest, whose guest() offers tools in every workspace. Offer pay() to pay for a workspace's model calls, classify(path) to say what a file is to you, and as a guest imports(path) to take a file through one of your tools. A product released for 0.3.0 runs in this process as a home instead (home().opener()); the app still attaches it, and pays through its opener's gateway.
  4. A guest ships its manifest, written from its capability registry with biocognia_intelligence.manifest.build(...): every tool names its permission, and a licensed one its entitlement, which grants() unlocks.
  5. Test a whole conversation with biocognia_intelligence.testing (guest_of, manifest_of) and spatialagents' FakeGateway, then check an install with biocognia doctor.

The full contract, with how licences and credits work, is docs/intelligence.md in the BioCognia platform repository.

Metadata

Release files for biocognia-intelligence 0.3.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for biocognia-intelligence 0.3.3
File Size Uploaded Provenance Actions
biocognia_intelligence-0.3.3.tar.gz 436.4 kB Without provenance Details for biocognia_intelligence-0.3.3.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for biocognia-intelligence 0.3.3
File Interpreter ABI Platform Actions
biocognia_intelligence-0.3.3-py3-none-any.whl Python 3 none any Details for biocognia_intelligence-0.3.3-py3-none-any.whl

Total release size: 530.6 kB

Release files / biocognia_intelligence-0.3.3.tar.gz

Download URL biocognia_intelligence-0.3.3.tar.gz
Size 436.4 kB
Tags Source
SHA-256 checksum
How to use checksums
bdf3224177e2ef52d79baf3c75bc0b174a65cc439f68be1e4f5e56432133483c
BLAKE2b-256 checksum
How to use checksums
1213e8ca3074c2c24cf9f4b63d3a5ef233e24a322cb8e1c1d0b977fd58fd84be
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.8.14

Release files / biocognia_intelligence-0.3.3-py3-none-any.whl

Download URL biocognia_intelligence-0.3.3-py3-none-any.whl
Size 94.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
982d11a2a83c0b0a6267ddab4bb1c807fef3aeca7f63cb18af1782eed0c927fb
BLAKE2b-256 checksum
How to use checksums
6250c591d98055e9c43a733f5cbee8a22575a814e382a9c426047ba7b678a321
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.8.14

Security for biocognia-intelligence 0.3.3

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page