Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
| Lux Capital Fund VII,Modal Labs,0.32 | ||
| Lux Capital Fund VII,Daytona,0.18 | ||
| Coatue Early Stage Fund III,Anthropic,0.84 | ||
| Coatue Early Stage Fund III,Cognition,0.58 |
There was a problem hiding this comment.
Demo CSV files duplicated at package root
Low Severity
Four CSV demo data files (client_indirect_exposure.csv, client_lp_positions.csv, lattice_iv_co_investors.csv, lattice_iv_portfolio.csv) are committed at the mito-ai/ package root in addition to the identical-named files in mito-ai/opto-demo/. The root-level copies have fewer columns (missing entry_round, entry_valuation_musd, post_money_valuation_musd) and are not referenced by any code. These look like leftover working-directory artifacts from developing the opto demo.
Additional Locations (2)
Reviewed by Cursor Bugbot for commit 46666d1. Configure here.
| * Distributed under the terms of the GNU Affero General Public License v3.0 License. | ||
| */ | ||
|
|
||
| @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500&display=swap'); |
There was a problem hiding this comment.
Google Fonts request fires for all Mito themes
Medium Severity
The @import url('https://fonts.googleapis.com/css2?...') in opto.css fires for every Mito theme user, not just Opto. theme.css unconditionally imports opto.css, and all three themes (Mito Light, Mito Dark, Opto) load the same CSS file via manager.loadCSS(style). This adds an unnecessary external network dependency and latency for Mito Light and Mito Dark users, and will cause timeouts in air-gapped or restricted-network environments.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 46666d1. Configure here.
| Lux Capital Fund VII,Modal Labs,0.32,Series A,180 | ||
| Lux Capital Fund VII,Daytona,0.18,Seed,35 | ||
| Coatue Early Stage Fund III,Anthropic,0.84,Series E,40000 | ||
| Coatue Early Stage Fund III,Cognition,0.58,Series A,2000 |
There was a problem hiding this comment.
Duplicate demo data in three separate directories
Low Severity
The opto demo data files are duplicated identically across mito-ai/opto-demo/ and mito-ai/demos/opto-demo/ (same schema, same data). This creates a maintenance burden where changes to the demo data need to be made in multiple places. One of these directories appears to be redundant.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 77ad6e8. Configure here.
|
|
||
| As a follow up, you can select some data in the default dataframe output and ask "Explain what a churned customer is vs a active customer" No newline at end of file | ||
| As a follow up, you can select some data in the default dataframe output and ask "Explain what a churned customer is vs a active customer" | ||
| >>>>>>> dev |
There was a problem hiding this comment.
Unresolved merge conflict markers committed in README
High Severity
The file contains unresolved git merge conflict markers (<<<<<<< HEAD, =======, >>>>>>> dev). The entire README is broken and will render as garbled text with conflict markers visible to anyone reading the documentation.
Reviewed by Cursor Bugbot for commit 63ce3f7. Configure here.
| .stMainBlockContainer {padding: 2rem 1rem 2rem 1rem;} | ||
| .stMainBlockContainer {padding: 2rem 4rem 2rem 4rem;} | ||
| max-width: 1400px; | ||
| margin: 0 auto; |
There was a problem hiding this comment.
Invalid CSS rules in Streamlit
Medium Severity
In the injected <style> block, max-width: 1400px and margin: 0 auto sit at the stylesheet root instead of inside a selector (for example .stMainBlockContainer), so browsers ignore them and the intended centered wide layout never applies.
Reviewed by Cursor Bugbot for commit bc712a9. Configure here.
| ) | ||
| # Load data from CSV files | ||
| subscriptions_df = pd.read_csv('subscriptions.csv') | ||
| customers_df = pd.read_csv('customers.csv') |
There was a problem hiding this comment.
Demo CSV paths need cwd
Medium Severity
The Streamlit demo now always calls pd.read_csv('subscriptions.csv') and pd.read_csv('customers.csv') relative to the process working directory. The previous file-upload flow worked from any launch directory; running the app elsewhere raises FileNotFoundError and the demo fails immediately.
Reviewed by Cursor Bugbot for commit bc712a9. Configure here.
| ) | ||
|
|
||
| sanitized_name = report_name.strip() | ||
| report = get_verified_report(sanitized_name) |
There was a problem hiding this comment.
Tool errors crash agent loop
High Severity
read_verified_report calls get_verified_report without handling failures from path sanitization (ValueError) or invalid JSON on disk (JSONDecodeError). Those propagate through _execute_tool, which has no surrounding try/except, and can abort the agent run instead of returning a failed ToolResult.
Reviewed by Cursor Bugbot for commit bc712a9. Configure here.
| "name": report.get("name", report_name), | ||
| "description": report.get("description", ""), | ||
| "snippet_count": len(snippets), | ||
| }) |
There was a problem hiding this comment.
Corrupt report breaks listing
Medium Severity
get_all_verified_reports wraps the directory walk in try/except OSError only, but _load_report can raise JSONDecodeError on a damaged .json file. One bad file then fails the whole listing, including system prompt formatting and “available reports” error text.
Reviewed by Cursor Bugbot for commit bc712a9. Configure here.
|
|
||
| if tool_result.tool_name == "scratchpad" and tool_result.success and tool_result.output: | ||
| content = tool_result.output | ||
| elif ( |
There was a problem hiding this comment.
Legacy assumptions break parsing
Medium Severity
analysis_assumptions is now a list of structured objects in AgentResponse, but parse_agent_response still passes LLM JSON straight into Pydantic with no coercion, so legacy string lists (or other shapes the frontend already normalizes) trigger validation errors and the malformed-response retry path.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 04aa497. Configure here.
| // Persisted instant-answer threads open the thread card | ||
| if (metadataThread) { | ||
| openThreadCard(app, notebookTracker, cell.model, metadataThread, rect); | ||
| return; |
There was a problem hiding this comment.
Output threads ignore extras
Medium Severity
Document-mode output comment indicators store only the first persisted thread per cell, so clicking the indicator always opens that thread even when additional output comment threads exist on the same cell.
Reviewed by Cursor Bugbot for commit de10027. Configure here.
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
There are 11 total unresolved issues (including 10 from previous reviews).
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit c352cfe. Configure here.
| else: | ||
| location = f"Cell {cell_number}, lines {start_line}-{end_line} (0 indexed)" | ||
| else: | ||
| location = f"Cell {cell_number} output" |
There was a problem hiding this comment.
Code thread mislabeled as output
Medium Severity
When formatting document_comment_thread context, a thread with type code but missing startLine is described as “Cell N output” instead of a code comment, so the agent can misread code discussions as output-only review threads.
Reviewed by Cursor Bugbot for commit c352cfe. Configure here.
1 participant
Description
Demo features :) Don't merge right now.
Testing
Documentation
Note if any new documentation needs to addressed or reviewed.
Note
Medium Risk
Touches agent response schema (
analysis_assumptions, new tool) and new persisted verified-report files/APIs; behavior is additive but incorrect assumptions UI could mislead users about analysis choices.Overview
Adds verified reports so teams can store annotated code snippets under
~/.mito/verified_reports/and the agent can load them via a newread_verified_reporttool, cite reused code withverified_snippet_ref/[MITO_VERIFIED_SNIPPET:…]in prompts, and surface reports through new REST routes undermito-ai/verified-reports.Agent & prompts:
analysis_assumptionsis now structured (selected,options,evidence) instead of plain strings. Chat context can includedocument_comment_threadso follow-up work can reference prior document review Q&A.Document review: New
comment_instant_answercompletion type streams short, ephemeral answers to code/output review comments without writing to chat history (runs in parallel with the agent).Demos & sample data: Fund-allocation demo (IPS vs portfolio, verified
fund-allocation-report.json, notebook + Streamlit app), retention demo/README and Streamlit tweaks (direct CSV load, layout), plus ancillary sample CSVs/notebooks for portfolio/LP scenarios.Reviewed by Cursor Bugbot for commit c352cfe. Bugbot is set up for automated code reviews on this repo. Configure here.