Skip to content

Cloud vs Local OSS

SpectraSherpa is one scientific platform with two practical entry points.

SpectraSherpa Cloud

Use Cloud when you want a browser-first managed environment with accounts, demo access, shared project context, and centrally configured AI features. Cloud is the right first path for an enterprise evaluator who wants to try FTIR, NIR, Raman, or UV-VIS workflows without installing Python.

The public demo is free to try at demo.spectrascientific.ai. Create an account with access code welcome_to_spectra_sherpa.

Cloud adds:

  • managed login and demo access codes
  • account-level limits for demo usage
  • Sherpa Advisor and Ambient Guidance
  • project memory and collaboration features
  • centrally managed model and LLM configuration

Cloud documentation is written for users, not developers. It explains how to use the hosted service and how to reason about its limits.

Local OSS

Use local OSS when you want to run SpectraSherpa on your own computer, inspect or modify the source, build plugins, or work without a hosted account.

Local OSS provides:

  • local FastAPI/Vue application
  • local SQLite-backed app data
  • workflow builder and node execution
  • core numpy/scipy/scikit-learn chemometrics
  • bring-your-own-key chat configuration through an OpenAI-compatible chat endpoint
  • optional spectra-sherpa[scp] support for SpectroChemPy-backed readers and nodes
  • optional spectra-sherpa[hitran] support for HITRAN/HAPI synthesis

AI Configuration and BYOK

SpectraSherpa separates scientific computation from AI assistance. PCA, PLS, SIMCA, KNN, MCR, preprocessing, and most plots run as deterministic application code. LLM features are optional assistance layers around interpretation, writing, workflow guidance, and selected suggestion tasks.

Local OSS can use a bring-your-own-key chat endpoint configured by environment variables. The built-in local chat path expects an OpenAI-compatible /chat/completions endpoint, which can be supplied by OpenAI, compatible hosted providers, or a local server that implements the same API shape. Anthropic/Claude access depends on the configured deployment path: Cloud deployments can be centrally configured for Claude, while local deployments need an OpenAI-compatible bridge or a provider path supported by the operator.

Cloud can be configured by Spectra Scientific for managed LLM access, BYOK access, or both, depending on the enterprise/demo profile. BYOK means your organization is also depending on the LLM supplier for model availability, usage limits, pricing, data-handling terms, and service reliability.

AI Use in Scientific Computing

Treat AI output as a labeled draft or suggestion, not a measurement. LLMs can be plausible and wrong, can miss sample-ID or validation-design problems, can overstate weak evidence, and can change behavior when an upstream provider updates a model.

AI-generated or AI-assisted artifacts in SpectraSherpa are labeled in the product. The main impacted features are:

  • Sherpa Advisor chat responses
  • Ambient Guidance suggestions
  • AI report narratives and data stories
  • peak-identification or vibrational-assignment suggestions when enabled
  • workflow or code-generation suggestions when enabled

Review AI output against spectra, metadata, file provenance, validation splits, plots, metrics, lab records, and domain knowledge before using it in scientific or customer-facing decisions.

The Main Difference

Cloud is for using the managed enterprise product. Local OSS is for self-hosted use, experimentation, and extension. The scientific workflow concepts are shared, but cloud-only account, demo-limit, collaboration, and centrally managed AI behavior is not part of the OSS package.