Skip to content

fix: [FEATURE] AI-powered dependency conflict prediction #554

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
jeffmarcilliat wants to merge 1 commit into from
Closed

fix: [FEATURE] AI-powered dependency conflict prediction #554

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
320 changes: 320 additions & 0 deletions cortex/cli.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -632,19 +632,307 @@ def ask(self, question: str) -> int:
self._print_error(str(e))
return 1

def _run_conflict_prediction(self, software: str) -> int | None:
"""
Run AI-powered conflict prediction before installation.

Returns:
None if safe to proceed
int (exit code) if should stop
"""
from cortex.conflict_predictor import (
ConflictSeverity,
DependencyConflictPredictor,
)

cx_print("Analyzing dependencies for potential conflicts...", "info")

try:
predictor = DependencyConflictPredictor()

# Split software into individual packages
packages = software.split()

# Run prediction for each package
all_conflicts = []
for pkg in packages:
# Skip commands like "pip install" or "apt-get"
if pkg in ("pip", "pip3", "apt", "apt-get", "install", "-y", "&&"):
continue

prediction = predictor.predict_conflicts(pkg)

if prediction.conflicts:
all_conflicts.append(prediction)

if not all_conflicts:
cx_print("No conflicts predicted. Safe to proceed.", "success")
console.print()
return None # Safe to proceed

# Display conflicts
console.print()
for prediction in all_conflicts:
# Format the prediction nicely
risk_emoji = {
ConflictSeverity.LOW: "[yellow]",
ConflictSeverity.MEDIUM: "[orange1]",
ConflictSeverity.HIGH: "[red]",
ConflictSeverity.CRITICAL: "[bold red]",
}
risk_color = risk_emoji.get(prediction.overall_risk, "")

console.print(
f"{risk_color}Conflict predicted[/]: {prediction.package_name}"
)
console.print(
f" Risk Level: {risk_color}{prediction.overall_risk.value.upper()}[/]"
)
console.print(
f" Confidence: {prediction.prediction_confidence:.0%}"
)
console.print()

# Show each conflict
for i, conflict in enumerate(prediction.conflicts, 1):
console.print(f" {i}. {conflict.description}")
console.print(
f" {conflict.conflicting_package} "
f"{conflict.conflicting_version} (installed)"
)
console.print()

# Show top suggestions
if prediction.resolutions:
console.print(" [bold cyan]Suggestions (ranked by safety):[/bold cyan]")
for i, res in enumerate(prediction.resolutions[:3], 1):
rec_tag = " [RECOMMENDED]" if res.recommended else ""
console.print(f" {i}. {res.description}{rec_tag}")
if res.command:
console.print(f" [dim]$ {res.command}[/dim]")
console.print()
Comment on lines +709 to +713
Copy link
Contributor

@coderabbitai coderabbitai bot Jan 10, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟡 Minor

Don’t double-print “[RECOMMENDED]”.

_run_conflict_prediction() adds rec_tag, but the predictor also mutates ResolutionSuggestion.description to include [RECOMMENDED] for the “safest” option, so you can end up with duplicated tags in output. Prefer leaving description unmodified and only styling in the CLI.

🤖 Prompt for AI Agents 
In @cortex/cli.py around lines 709 - 713, The CLI is double-printing "
[RECOMMENDED]" because _run_conflict_prediction() appends rec_tag when printing
but the predictor also mutates ResolutionSuggestion.description to include "
[RECOMMENDED]"; remove the mutation inside the predictor (stop modifying
ResolutionSuggestion.description) and let the CLI add rec_tag based on
res.recommended (in _run_conflict_prediction()), ensuring the predictor sets
only the boolean flag (ResolutionSuggestion.recommended) and not the description
text.

Comment on lines +675 to +713
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Jan 10, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The code for displaying conflict predictions here is very similar to the code in the predict method (lines 810-874). This duplication can make future changes more difficult, as they would need to be applied in two places.

Consider refactoring this display logic into a separate helper function that takes a prediction object and handles the formatting and printing. This would improve maintainability and reduce code duplication.


# Check if we should stop for critical conflicts
critical_predictions = [
p for p in all_conflicts if p.overall_risk == ConflictSeverity.CRITICAL
]

if critical_predictions:
cx_print(
"Critical conflicts detected. Installation blocked.",
"error",
)
cx_print(
"Resolve conflicts above or use --no-predict to skip this check.",
"warning",
)
return 1 # Stop with error

# For non-critical conflicts, ask user
high_predictions = [
p for p in all_conflicts if p.overall_risk == ConflictSeverity.HIGH
]

if high_predictions:
try:
response = console.input(
"[bold yellow]High-risk conflicts detected. "
"Proceed anyway? (y/N): [/bold yellow]"
)
if response.lower() not in ("y", "yes"):
cx_print("Installation cancelled", "info")
return 0 # User cancelled
except (EOFError, KeyboardInterrupt):
console.print()
cx_print("Installation cancelled", "info")
return 0

# Medium/Low conflicts - just warn and proceed
return None # Safe to proceed

except ImportError as e:
# Conflict predictor module not available - just warn and continue
self._debug(f"Conflict prediction unavailable: {e}")
return None
except Exception as e:
# Don't let prediction errors block installation
cx_print(f"Conflict prediction failed: {e}", "warning")
cx_print("Proceeding with installation...", "info")
return None

Comment on lines +635 to +762
Copy link
Contributor

@coderabbitai coderabbitai bot Jan 10, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟡 Minor

Make --predict and --no-predict mutually exclusive (and define precedence).

Right now users can pass both; argparse will accept it and your should_predict logic will silently prefer --predict. Use a mutually-exclusive group for install flags.

Also applies to: 913-935, 2426-2459

def predict(self, package: str, json_output: bool = False, verbose: bool = False) -> int:
"""
Predict dependency conflicts for a package before installation.

This is the standalone 'cortex predict' command.
"""
import json as json_lib

from cortex.conflict_predictor import (
ConflictSeverity,
DependencyConflictPredictor,
)

try:
predictor = DependencyConflictPredictor()

# Split packages if multiple provided
packages = package.split()
predictions = []

cx_print(f"Analyzing {len(packages)} package(s) for conflicts...", "info")
console.print()

for pkg in packages:
prediction = predictor.predict_conflicts(pkg)
predictions.append(prediction)

if json_output:
# JSON output mode
output = {
"packages": [
predictor.export_prediction_json(p) for p in predictions
],
"summary": {
"total_packages": len(predictions),
"packages_with_conflicts": sum(
1 for p in predictions if p.conflicts
),
"critical_conflicts": sum(
1
for p in predictions
if p.overall_risk == ConflictSeverity.CRITICAL
),
},
}
console.print(json_lib.dumps(output, indent=2))
else:
# Human-readable output
for prediction in predictions:
if not prediction.conflicts:
console.print(
f"[green]No conflicts predicted[/green] for "
f"[bold]{prediction.package_name}[/bold]"
)
console.print(
f" Confidence: {prediction.prediction_confidence:.0%}"
)
console.print()
continue

# Risk color coding
risk_color = {
ConflictSeverity.LOW: "yellow",
ConflictSeverity.MEDIUM: "orange1",
ConflictSeverity.HIGH: "red",
ConflictSeverity.CRITICAL: "bold red",
}.get(prediction.overall_risk, "white")

console.print(
f"[{risk_color}]Conflict predicted[/{risk_color}]: "
f"[bold]{prediction.package_name}[/bold]"
)
console.print()

# Show conflicts
for i, conflict in enumerate(prediction.conflicts, 1):
severity_badge = f"[{conflict.severity.value.upper()}]"
console.print(f" {i}. {severity_badge} {conflict.description}")
console.print(
f" Installed: {conflict.conflicting_package} "
f"{conflict.conflicting_version}"
)
console.print(f" Type: {conflict.conflict_type}")
console.print(
f" Confidence: {conflict.confidence:.0%}"
)
console.print()

# Show suggestions
if prediction.resolutions:
console.print(
" [bold cyan]Suggestions (ranked by safety):[/bold cyan]"
)
for i, res in enumerate(prediction.resolutions[:4], 1):
rec = " [RECOMMENDED]" if res.recommended else ""
console.print(f" {i}. {res.description}")
if res.command:
console.print(f" [dim]$ {res.command}[/dim]")
console.print(
f" Safety: {res.safety_score:.0%}{rec}"
)
if verbose and res.side_effects:
for effect in res.side_effects:
console.print(f" [dim]- {effect}[/dim]")
console.print()

# Show analysis details if verbose
if verbose and prediction.analysis_details:
console.print(" [dim]Analysis Details:[/dim]")
for key, value in prediction.analysis_details.items():
console.print(f" {key}: {value}")
console.print()

# Summary
conflicts_found = sum(1 for p in predictions if p.conflicts)
critical = sum(
1
for p in predictions
if p.overall_risk == ConflictSeverity.CRITICAL
)

console.print("-" * 50)
if conflicts_found == 0:
console.print(
f"[bold green]All {len(predictions)} package(s) safe to install[/bold green]"
)
return 0
else:
console.print(
f"[bold yellow]{conflicts_found}/{len(predictions)} package(s) "
f"have potential conflicts[/bold yellow]"
)
if critical > 0:
console.print(
f"[bold red]{critical} CRITICAL conflict(s) detected[/bold red]"
)
return 1
return 0

except ImportError as e:
self._print_error(f"Conflict prediction module not available: {e}")
return 1
except Exception as e:
self._print_error(f"Prediction failed: {e}")
if verbose:
import traceback

traceback.print_exc()
return 1

def install(
self,
software: str,
execute: bool = False,
dry_run: bool = False,
parallel: bool = False,
predict: bool = False,
no_predict: bool = False,
):
# Validate input first
is_valid, error = validate_install_request(software)
if not is_valid:
self._print_error(error)
return 1

# Run conflict prediction if requested or if executing (unless --no-predict)
should_predict = predict or (execute and not no_predict)
if should_predict:
prediction_result = self._run_conflict_prediction(software)
if prediction_result is not None:
# Non-None means we should stop (critical conflict or user cancelled)
return prediction_result

# Special-case the ml-cpu stack:
# The LLM sometimes generates outdated torch==1.8.1+cpu installs
# which fail on modern Python. For the "pytorch-cpu jupyter numpy pandas"
Expand Down Expand Up @@ -2031,6 +2319,7 @@ def show_rich_help():
table.add_row("wizard", "Configure API key")
table.add_row("status", "System status")
table.add_row("install <pkg>", "Install software")
table.add_row("predict <pkg>", "Predict dependency conflicts")
table.add_row("import <file>", "Import deps from package files")
table.add_row("history", "View history")
table.add_row("rollback <id>", "Undo installation")
Expand Down Expand Up @@ -2144,6 +2433,29 @@ def main():
action="store_true",
help="Enable parallel execution for multi-step installs",
)
install_parser.add_argument(
"--predict",
action="store_true",
help="Predict dependency conflicts before installation",
)
install_parser.add_argument(
"--no-predict",
action="store_true",
help="Skip automatic conflict prediction",
)

# Predict command - standalone conflict prediction
predict_parser = subparsers.add_parser(
"predict",
help="Predict dependency conflicts before installation",
)
predict_parser.add_argument("package", type=str, help="Package(s) to analyze")
predict_parser.add_argument(
"--json", "-j", action="store_true", help="Output as JSON"
)
predict_parser.add_argument(
"--verbose", action="store_true", help="Show detailed analysis"
)

# Import command - import dependencies from package manager files
import_parser = subparsers.add_parser(
Expand Down Expand Up @@ -2510,6 +2822,14 @@ def main():
execute=args.execute,
dry_run=args.dry_run,
parallel=args.parallel,
predict=args.predict,
no_predict=args.no_predict,
)
elif args.command == "predict":
return cli.predict(
args.package,
json_output=args.json,
verbose=args.verbose,
)
elif args.command == "import":
return cli.import_deps(args)
Expand Down
Loading
Loading