For the complete documentation index, see /llms.txt. The full corpus is available at /llms-full.txt. Most pages have a raw markdown variant; append .md to the URL or send Accept: text/markdown.
PricingComplianceCareersBook a demo

Docs  |  User Guides

Assist

Invert Assist lets you query your bioprocess data in plain language. Instead of navigating manually through plots and exports, you can ask questions conversationally and receive results generated directly from your data. Behind the scenes, Assist writes and executes Python code against your data, so every answer is reproducible and traceable.

What you can do with Invert Assist

  • Outlier Detection: Scan all your timeseries data to surface excursions worth investigating further.
  • Root Cause Analysis: Identify drivers behind unexpected trends or deviations in your process data.
  • Experiment Summarization: Generate clear summaries of multi-run experiments, highlighting key similarities and differences.
  • Iterative Exploration: Refine questions and follow up to dig deeper — Assist retains context across the full conversation.
  • Report Generation: Ask Assist to create comprehensive analysis reports with charts, tables, and summaries — automatically structured and ready for review.
  • Product Q&A: Ask Assist how Invert features work — it can consult the Invert product documentation to answer questions about the platform.

Generating Reports with Assist

Assist can create complete analysis reports from your bioprocess data. Instead of manually building reports from charts and tables, you can ask Assist to generate one in plain language.

How to Generate a Report

  1. Select your runs - Choose the runs you want analyzed
  2. Open Assist - Click the Assist button at the bottom of the screen
  3. Ask for a report - Try prompts like:
    • "Generate a process monitoring report for my most recent experiment"
    • "Create a comparative analysis of these two runs"
    • "Build a summary report highlighting key metrics and trends"
  4. Review the result - Assist will produce a report with analysis, charts, and data tables
  5. Refine if needed - Ask Assist to modify specific sections or regenerate with a different prompt
  6. Save to Reports - Once satisfied, save the report to your Reports library

Tips for Better Reports

  • Be specific about what analysis you want (e.g., "Compare yield between these two strains")
  • Mention timeframes if relevant (e.g., "Report on the growth phase data")
  • Ask follow-ups to refine the analysis (e.g., "Can you add a section on pH impact?")
  • Review for accuracy before saving, especially for presentation or publication use

How to use Invert Assist

  1. Select your runs. For best performance, starting with a focused set of runs (around 15 or fewer) produces the fastest, most targeted results — you can always load additional runs mid-conversation using Assist's run loading tools.

  2. Open the Invert Assist chat panel using the Assist button at the bottom of your screen:

  3. Type your question—for example:

    • “What caused the excursion in Run75?"
    • "Is there any effect of pH on titer based on these runs?"
    • "What is the next experiment that might be interesting to explore?"

  4. Review the answer and trace the reasoning chain alongside the code that was executed for the analysis.

  5. Use follow-up questions to refine your results or switch context.

History and context in Assist

  • Your Assist queries and outputs are saved at a user-level. To view a repository of past chats, click the clock icon in the Assist modal.

  • In addition to Runs, Notebooks can also be provided as context to Assist. This allows users to leverage existing analysis templates as a reference for Assist to perform calculations. For this, navigate to the relevant notebook page, open Assist modal, and add notebook as context.

  • Workspace admins can further add context at the organizational level, which will be used in all Assist queries for your team. Use this space to provide terminology, conventions, or guidelines that should inform the assistant's responses. Organization context can be added under Settings > AI in the left navigation.

Editing Reports with Assist

For organizations with the Assist report editing feature enabled, Assist can also modify existing saved reports — not just generate new ones. Open a report, launch Assist, and ask it to update specific sections, add new plot blocks, or revise text.

When Assist edits a report, it produces a draft — an unsaved, private suggestion you can review before saving. Drafts have these characteristics:

  • Unsaved: Drafts are private to you until you explicitly save them
  • Editable: You can accept the draft as-is or ask Assist to modify it further before saving
  • Stale draft warning: If a draft sits for more than 5 minutes without being reviewed, you'll see a warning (⚠️) indicating the analysis may be outdated
  • Accept or reject: You control whether the draft is applied to the saved report or discarded
  • Visual diff on review: When you open a draft, modified blocks are marked with a blue stripe and added blocks with a green stripe so you can spot Assist's changes at a glance.

Tips for better results

  • Be specific: include timeframes or phases, run names, or metrics in your question.
  • Use follow-ups: instead of repeating full queries, build on the last response.

FAQs

Will Invert Assist chat change my data?

No. Invert Assist's chat tooling is read-only. It queries and analyzes data but never edits it.

What data can Invert Assist chat access?

It has access to your structured bioprocess data stored in Invert, including runs, metrics, properties, events, lineage, formulas, and reports. Your data is processed within Amazon Bedrock infrastructure and is never used for model training or fine-tuning.

How does Invert ensure reliability?

Invert maintains a suite of automated benchmarking evaluations that run continuously to characterize Assist's analytical capabilities and detect any regressions caused by model or infrastructure changes.

What if the result is incorrect or Assist misunderstands the question?

Try rephrasing with more specific detail — for example, naming the run, metric, or time range you have in mind. You can also submit feedback using the thumbs-up and thumbs-down icons after a response, which helps the team investigate edge cases and improve accuracy over time.

Still need help? Reach out to our support team via the Help & Support link in the app.

Data QualitySkills