Nl parser #544
Nl parser #544
Conversation
…53/cortex into nl-parser-feature # Conflicts: # cortex/cli.py
…53/cortex into nl-parser-feature
📝 WalkthroughWalkthroughThis pull request adds a natural language installation feature enabling users to install software via conversational input. Changes introduce intent extraction across multiple LLM providers (OpenAI, Claude, Ollama), interactive clarification workflows for ambiguous requests, stdin-aware prompt construction, and virtual environment command normalization. The CommandInterpreter class gains intent extraction and natural language understanding methods; the CortexCLI class incorporates ambiguity detection and interactive decision flows into the install workflow. Legacy shell-environment-analyzer subcommands are removed. Changes
Sequence Diagram(s)sequenceDiagram
participant User
participant CLI as CortexCLI
participant Interp as CommandInterpreter
participant LLM as LLM API
User->>CLI: install("install pytorch")
CLI->>CLI: _build_prompt_with_stdin(user_input)
CLI->>Interp: extract_intent(prompt)
Interp->>LLM: POST intent extraction request
LLM-->>Interp: intent JSON response
Interp->>Interp: _parse_intent_from_text(response)
Interp-->>CLI: intent dict
alt is_ambiguous_request
CLI->>CLI: _clarification_prompt(user_input, intent)
CLI->>Interp: ask_clarifying_questions(user_input, intent)
Interp->>LLM: POST clarifying questions request
LLM-->>Interp: questions
Interp-->>CLI: questions list
CLI->>CLI: _generate_suggestions(interpreter, user_input, intent)
CLI->>LLM: POST suggestion generation
LLM-->>CLI: command suggestions
CLI->>User: Display suggestions, prompt for selection/edit
User->>CLI: User selection/confirmation
CLI->>Interp: refine_intent(original_request, clarification)
Interp-->>CLI: refined intent
else not ambiguous
CLI->>CLI: Proceed with direct command generation
end
CLI->>Interp: get_commands(refined_intent)
Interp-->>CLI: command list
CLI->>CLI: _normalize_venv_commands(commands)
CLI->>User: Display commands (dry-run or execute)
alt execute mode
User->>CLI: Confirm execution
CLI->>CLI: Execute commands
else dry_run mode
CLI->>User: Show commands without execution
end
sequenceDiagram
participant Interp as CommandInterpreter
participant OpenAI as OpenAI API
participant Claude as Claude API
participant Ollama as Ollama API
Interp->>Interp: extract_intent(user_input)
alt provider == OPENAI
Interp->>OpenAI: _call_openai(_get_intent_prompt())
OpenAI-->>Interp: JSON response
Interp->>Interp: _parse_intent_from_text(response)
else provider == CLAUDE
Interp->>Claude: _call_claude(_get_intent_prompt())
Claude-->>Interp: text response
Interp->>Interp: _parse_intent_from_text(response)
else provider == OLLAMA
Interp->>Interp: _get_ollama_model()
Interp->>Ollama: POST request (base_url + model)
Ollama-->>Interp: JSON response
Interp->>Interp: _parse_intent_from_text(response)
else provider == FAKE
Interp->>Interp: Return configurable/default intent
end
Interp-->>Interp: return intent dict
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested labels
Suggested reviewers
Poem
🚥 Pre-merge checks | ❌ 3❌ Failed checks (2 warnings, 1 inconclusive)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
CLA Verification PassedAll contributors have signed the CLA.
|
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 7
🤖 Fix all issues with AI agents
In @cortex/cli.py:
- Around line 75-78: The inline comment in cortex/cli.py describing the
confidence threshold is misleading: it says "lowered from 0.5 to be more
conservative" while the code uses 0.7 (which is a raise). Update the comment
near the ambiguous-check block (the lines describing domain unknown, confidence
< 0.7, and short requests) to correctly state that the confidence threshold was
raised/increased from 0.5 to 0.7 (or simply say "threshold increased to 0.7 for
conservatism") so the comment matches the actual logic in that function/block.
- Around line 130-179: The provider comparisons in the chat branch use
interpreter.provider.name (which yields enum member names like "OPENAI") causing
the wrong path to be taken; switch comparisons to use interpreter.provider.value
or compare against the APIProvider enum members directly (e.g.,
APIProvider.OPENAI) and update all occurrences in the function handling the
provider logic (the blocks referencing interpreter.provider.name,
interpreter.client, interpreter.model, and interpreter.ollama_url) so each
provider-specific branch (OpenAI/Claude/Ollama/Fake) is matched correctly.
In @cortex/llm/interpreter.py:
- Around line 575-604: Add a proper docstring to the public method
extract_intent describing its purpose (parse user input into an intent dict),
the parameters (user_input: str), the return type (dict with keys like action,
domain, install_mode, description, ambiguous, confidence and that it may vary by
provider), and raised exceptions (ValueError for empty input or unsupported
provider, JSONDecodeError handling for CORTEX_FAKE_INTENT). Mention supported
providers (APIProvider.OPENAI, APIProvider.CLAUDE, APIProvider.OLLAMA,
APIProvider.FAKE) and note that when using FAKE it may read CORTEX_FAKE_INTENT
from the environment as JSON fallback behavior.
- Around line 328-335: The fallback intent returned in the parsing-failure
branch is missing the install_mode field; update the returned dict in the
failure branch (the anonymous intent fallback returned in interpreter.py) to
include "install_mode" with the same sentinel value used by other fallbacks
(e.g., "unknown" or the project’s standard fallback value) so the structure
matches other intent fallback objects.
- Around line 233-242: The JSON example in the prompt block under the Format:
docstring in cortex/llm/interpreter.py has a missing comma after the
"install_mode": "..." entry; update that prompt/example (the triple-quoted
string starting with "Format:") to include a comma after "install_mode": "..."
so the example becomes valid JSON and won't confuse the LLM.
- Around line 194-202: The Ollama fallback returned in the except Exception
block in cortex/llm/interpreter.py is missing the install_mode key, which causes
downstream KeyError when code reads intent.get("install_mode"); update that
returned dict (the one currently returning
action/domain/description/ambiguous/confidence) to include "install_mode":
"unknown" (matching the shape used by the OpenAI/Claude fallbacks) so the intent
structure is consistent.
In @docs/docs/nl-installer.md:
- Around line 1-18: Update the incorrect test file reference in the NL Installer
docs: change the mention of tests/test_nl_installer.py to the correct
tests/test_nl_parser_cases.py in the docs/docs/nl-installer.md content (the
paragraph that currently states "The NL installer is validated using unit tests
in `tests/test_nl_installer.py`"). Ensure the updated line reads that the NL
installer is validated using unit tests in `tests/test_nl_parser_cases.py`.
🧹 Nitpick comments (5)
cortex/llm/interpreter.py (4)
72-94: Add explicit encoding and consider narrowing exception handling.Two minor improvements:
- Line 85: Add
encoding="utf-8"toopen()for cross-platform consistency- Line 90: The bare
except Exceptionis broad; consider catching(OSError, json.JSONDecodeError, KeyError)for claritySuggested improvement
config_file = Path.home() / ".cortex" / "config.json" if config_file.exists(): - with open(config_file) as f: + with open(config_file, encoding="utf-8") as f: config = json.load(f) model = config.get("ollama_model") if model: return model - except Exception: + except (OSError, json.JSONDecodeError, KeyError): pass # Ignore errors reading config
188-202: Narrow the exception handling to avoid catching system exceptions.The bare
except Exceptionon line 194 could catchKeyboardInterruptorSystemExit. Consider catching specific network-related exceptions.Suggested fix
- except Exception: + except (urllib.error.URLError, json.JSONDecodeError, OSError, TimeoutError) as e: # True failure → unknown intent return { "action": "unknown", "domain": "unknown", - "description": "Failed to extract intent", + "description": f"Failed to extract intent: {e}", "ambiguous": True, "confidence": 0.0, }
678-680: Avoid bareexceptclause.The bare
excepton line 678 will catch all exceptions includingKeyboardInterruptandSystemExit. Consider catching specific exceptions or at minimum useexcept Exception:.
752-755: Add docstring for public method.The
refine_intentmethod is a public API and should have a docstring per coding guidelines.Add docstring
def refine_intent(self, original_request: str, clarification: str) -> dict: - """Refine intent using a user clarification, delegating back to extraction.""" + """Refine intent by combining original request with user clarification. + + Args: + original_request: The initial user request + clarification: Additional context provided by the user + + Returns: + Updated intent dict with refined understanding + """ combined = f"{original_request}\nClarification: {clarification}" return self.extract_intent(combined)cortex/cli.py (1)
194-196: Unused exception variable and missing debug logging.The exception is caught but not logged despite the comment indicating debug logging was intended. Consider implementing the debug logging or removing the misleading comment.
Add debug logging
except Exception as e: - # Log error in debug mode - pass + # Silent fallback - suggestions are non-critical + # Debug: print(f"Suggestion generation failed: {e}", file=sys.stderr) + pass
📜 Review details
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
cortex/cli.pycortex/llm/interpreter.pydocs/docs/nl-installer.mdtests/test_nl_parser_cases.py
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit inference engine (AGENTS.md)
**/*.py: Follow PEP 8 style guide
Type hints required in Python code
Docstrings required for all public APIs
Files:
tests/test_nl_parser_cases.pycortex/llm/interpreter.pycortex/cli.py
tests/**/*.py
📄 CodeRabbit inference engine (AGENTS.md)
Maintain >80% test coverage for pull requests
Files:
tests/test_nl_parser_cases.py
🧠 Learnings (4)
📓 Common learnings
Learnt from: pavanimanchala53
Repo: cortexlinux/cortex PR: 439
File: cortex/cli.py:62-73
Timestamp: 2026-01-05T20:26:18.626Z
Learning: In cortex/cli.py, the _is_ambiguous_request method intentionally only checks for domain=="unknown" and does not recheck the ambiguous field from the intent dict. The design assumes that ambiguity handling occurs at the interpreter layer through LLM prompting, and the domain check in CLI is sufficient as a safety mechanism.
📚 Learning: 2026-01-05T20:26:18.626Z
Learnt from: pavanimanchala53
Repo: cortexlinux/cortex PR: 439
File: cortex/cli.py:62-73
Timestamp: 2026-01-05T20:26:18.626Z
Learning: In cortex/cli.py, the _is_ambiguous_request method intentionally only checks for domain=="unknown" and does not recheck the ambiguous field from the intent dict. The design assumes that ambiguity handling occurs at the interpreter layer through LLM prompting, and the domain check in CLI is sufficient as a safety mechanism.
Applied to files:
cortex/llm/interpreter.pycortex/cli.py
📚 Learning: 2025-12-11T12:03:24.071Z
Learnt from: CR
Repo: cortexlinux/cortex PR: 0
File: AGENTS.md:0-0
Timestamp: 2025-12-11T12:03:24.071Z
Learning: Applies to **/*install*.py : Dry-run by default for all installations in command execution
Applied to files:
cortex/cli.py
📚 Learning: 2025-12-11T12:03:24.071Z
Learnt from: CR
Repo: cortexlinux/cortex PR: 0
File: AGENTS.md:0-0
Timestamp: 2025-12-11T12:03:24.071Z
Learning: Applies to **/*install*.py : No silent sudo execution - require explicit user confirmation
Applied to files:
cortex/cli.py
🧬 Code graph analysis (1)
tests/test_nl_parser_cases.py (1)
cortex/llm/interpreter.py (2)
parse(505-563)extract_intent(575-604)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: test (3.12)
- GitHub Check: test (3.10)
- GitHub Check: test (3.11)
🔇 Additional comments (11)
cortex/llm/interpreter.py (2)
42-44: Good defensive initialization, minor redundancy noted.The Ollama base URL is correctly initialized here, but note that
_initialize_client(line 116) also reads from the same environment variable. Consider storingself.ollama_urlunconditionally in__init__and reusing it in_initialize_clientto avoid the duplicateos.environ.getcall.
692-750: LGTM - good defensive parsing of LLM responses.The filtering logic on lines 741-746 effectively removes meta-commentary from LLM responses. The silent exception handling on line 749-750 is acceptable for a non-critical feature, though consider adding debug logging if verbose mode is available.
tests/test_nl_parser_cases.py (4)
8-19: Well-designed test fixture.The fixture properly isolates tests from external API calls by using the fake provider with predefined responses.
22-80: Tests validate parsing doesn't crash, but have limited verification depth.These tests effectively verify that the
parsemethod handles various input types without errors. However, since the fake provider returns identical commands regardless of input, consider adding tests withCORTEX_FAKE_INTENTto verify that different intents are correctly extracted for different request types.
127-159: Good coverage of fake intent injection mechanism.These tests properly verify that the
CORTEX_FAKE_INTENTenvironment variable correctly overrides the default fake intent, enabling controlled testing of downstream intent-handling logic.
82-88: Good test documenting default fake provider behavior.This test ensures that the default fake intent is non-ambiguous, which is important for downstream test isolation.
cortex/cli.py (5)
38-52: Good stdin handling with defensive error catching.The stdin consumption in
__init__correctly handles the case where stdin is not a tty and gracefully handlesOSError.
283-316: Good platform-aware venv command normalization.The implementation correctly detects venv creation and normalizes activation commands for both Windows and Unix platforms.
800-842: Well-implemented interactive clarification loop.The clarification flow correctly:
- Limits rounds to prevent infinite loops (max 3)
- Handles both question-based and free-form clarification
- Fails fast in non-interactive mode with helpful suggestions
Based on learnings, the ambiguity check at the CLI layer using domain is intentional as a safety mechanism.
896-944: Good confirmation flow with edit capability.The confirmation flow provides flexibility with yes/edit/no options. The auto-approve in non-interactive mode (line 900-901) is necessary for CI/testing and is appropriately flagged with a warning. This aligns with the learning that execution requires the explicit
--executeflag.
856-875: Good install mode-aware prompt construction.The prompt correctly adjusts based on
install_mode- using pip/venv for Python packages and apt for system packages. The command normalization (line 875) ensures venv activation is platform-appropriate.
| # Consider ambiguous if: | ||
| # - domain unknown | ||
| # - confidence < 0.7 (lowered from 0.5 to be more conservative) | ||
| # - request is very short (< 5 words) suggesting underspecification |
There was a problem hiding this comment.
Comment says "lowered" but threshold was raised.
Line 77 states the confidence threshold was "lowered from 0.5 to be more conservative" but 0.7 > 0.5, meaning it was actually raised to be more conservative (requiring higher confidence). This is correct behavior but the comment is misleading.
Fix comment
# Consider ambiguous if:
# - domain unknown
- # - confidence < 0.7 (lowered from 0.5 to be more conservative)
+ # - confidence < 0.7 (raised from 0.5 to be more conservative)
# - request is very short (< 5 words) suggesting underspecification📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| # Consider ambiguous if: | |
| # - domain unknown | |
| # - confidence < 0.7 (lowered from 0.5 to be more conservative) | |
| # - request is very short (< 5 words) suggesting underspecification | |
| # Consider ambiguous if: | |
| # - domain unknown | |
| # - confidence < 0.7 (raised from 0.5 to be more conservative) | |
| # - request is very short (< 5 words) suggesting underspecification |
🤖 Prompt for AI Agents
In @cortex/cli.py around lines 75 - 78, The inline comment in cortex/cli.py
describing the confidence threshold is misleading: it says "lowered from 0.5 to
be more conservative" while the code uses 0.7 (which is a raise). Update the
comment near the ambiguous-check block (the lines describing domain unknown,
confidence < 0.7, and short requests) to correctly state that the confidence
threshold was raised/increased from 0.5 to 0.7 (or simply say "threshold
increased to 0.7 for conservatism") so the comment matches the actual logic in
that function/block.
| if interpreter.provider.name == "openai": | ||
| response = interpreter.client.chat.completions.create( | ||
| model=interpreter.model, | ||
| messages=[ | ||
| { | ||
| "role": "system", | ||
| "content": "You are a helpful assistant that suggests installation requests. Be specific and relevant.", | ||
| }, | ||
| {"role": "user", "content": prompt}, | ||
| ], | ||
| temperature=0.3, | ||
| max_tokens=200, | ||
| ) | ||
| content = response.choices[0].message.content.strip() | ||
| elif interpreter.provider.name == "claude": | ||
| response = interpreter.client.messages.create( | ||
| model=interpreter.model, | ||
| max_tokens=200, | ||
| temperature=0.3, | ||
| system="You are a helpful assistant that suggests installation requests. Be specific and relevant.", | ||
| messages=[{"role": "user", "content": prompt}], | ||
| ) | ||
| content = response.content[0].text.strip() | ||
| elif interpreter.provider.name == "ollama": | ||
| full_prompt = f"System: You are a helpful assistant that suggests installation requests. Be specific and relevant.\n\nUser: {prompt}" | ||
| data = json.dumps( | ||
| { | ||
| "model": interpreter.model, | ||
| "prompt": full_prompt, | ||
| "stream": False, | ||
| "options": {"temperature": 0.3}, | ||
| } | ||
| ).encode("utf-8") | ||
| req = urllib.request.Request( | ||
| f"{interpreter.ollama_url}/api/generate", | ||
| data=data, | ||
| headers={"Content-Type": "application/json"}, | ||
| ) | ||
| with urllib.request.urlopen(req, timeout=30) as response: | ||
| result = json.loads(response.read().decode("utf-8")) | ||
| content = result.get("response", "").strip() | ||
| elif interpreter.provider.name == "fake": | ||
| # Return fake suggestions for testing | ||
| return [ | ||
| f"install {user_input} with more details", | ||
| f"set up {user_input} environment", | ||
| f"configure {user_input} tools", | ||
| ] | ||
| else: | ||
| return [] |
There was a problem hiding this comment.
Bug: Provider comparison uses wrong attribute.
The code compares interpreter.provider.name with lowercase strings like "openai", but APIProvider is an Enum where .name returns the uppercase member name (e.g., "OPENAI"), not the value. This causes all provider-specific paths to be skipped, always falling back to the default suggestions.
Use interpreter.provider.value or compare with the enum members directly.
Fix provider comparisons
- if interpreter.provider.name == "openai":
+ if interpreter.provider.value == "openai":
response = interpreter.client.chat.completions.create(
...
)
content = response.choices[0].message.content.strip()
- elif interpreter.provider.name == "claude":
+ elif interpreter.provider.value == "claude":
response = interpreter.client.messages.create(
...
)
content = response.content[0].text.strip()
- elif interpreter.provider.name == "ollama":
+ elif interpreter.provider.value == "ollama":
...
- elif interpreter.provider.name == "fake":
+ elif interpreter.provider.value == "fake":
# Return fake suggestions for testing
return [...]📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| if interpreter.provider.name == "openai": | |
| response = interpreter.client.chat.completions.create( | |
| model=interpreter.model, | |
| messages=[ | |
| { | |
| "role": "system", | |
| "content": "You are a helpful assistant that suggests installation requests. Be specific and relevant.", | |
| }, | |
| {"role": "user", "content": prompt}, | |
| ], | |
| temperature=0.3, | |
| max_tokens=200, | |
| ) | |
| content = response.choices[0].message.content.strip() | |
| elif interpreter.provider.name == "claude": | |
| response = interpreter.client.messages.create( | |
| model=interpreter.model, | |
| max_tokens=200, | |
| temperature=0.3, | |
| system="You are a helpful assistant that suggests installation requests. Be specific and relevant.", | |
| messages=[{"role": "user", "content": prompt}], | |
| ) | |
| content = response.content[0].text.strip() | |
| elif interpreter.provider.name == "ollama": | |
| full_prompt = f"System: You are a helpful assistant that suggests installation requests. Be specific and relevant.\n\nUser: {prompt}" | |
| data = json.dumps( | |
| { | |
| "model": interpreter.model, | |
| "prompt": full_prompt, | |
| "stream": False, | |
| "options": {"temperature": 0.3}, | |
| } | |
| ).encode("utf-8") | |
| req = urllib.request.Request( | |
| f"{interpreter.ollama_url}/api/generate", | |
| data=data, | |
| headers={"Content-Type": "application/json"}, | |
| ) | |
| with urllib.request.urlopen(req, timeout=30) as response: | |
| result = json.loads(response.read().decode("utf-8")) | |
| content = result.get("response", "").strip() | |
| elif interpreter.provider.name == "fake": | |
| # Return fake suggestions for testing | |
| return [ | |
| f"install {user_input} with more details", | |
| f"set up {user_input} environment", | |
| f"configure {user_input} tools", | |
| ] | |
| else: | |
| return [] | |
| if interpreter.provider.value == "openai": | |
| response = interpreter.client.chat.completions.create( | |
| model=interpreter.model, | |
| messages=[ | |
| { | |
| "role": "system", | |
| "content": "You are a helpful assistant that suggests installation requests. Be specific and relevant.", | |
| }, | |
| {"role": "user", "content": prompt}, | |
| ], | |
| temperature=0.3, | |
| max_tokens=200, | |
| ) | |
| content = response.choices[0].message.content.strip() | |
| elif interpreter.provider.value == "claude": | |
| response = interpreter.client.messages.create( | |
| model=interpreter.model, | |
| max_tokens=200, | |
| temperature=0.3, | |
| system="You are a helpful assistant that suggests installation requests. Be specific and relevant.", | |
| messages=[{"role": "user", "content": prompt}], | |
| ) | |
| content = response.content[0].text.strip() | |
| elif interpreter.provider.value == "ollama": | |
| full_prompt = f"System: You are a helpful assistant that suggests installation requests. Be specific and relevant.\n\nUser: {prompt}" | |
| data = json.dumps( | |
| { | |
| "model": interpreter.model, | |
| "prompt": full_prompt, | |
| "stream": False, | |
| "options": {"temperature": 0.3}, | |
| } | |
| ).encode("utf-8") | |
| req = urllib.request.Request( | |
| f"{interpreter.ollama_url}/api/generate", | |
| data=data, | |
| headers={"Content-Type": "application/json"}, | |
| ) | |
| with urllib.request.urlopen(req, timeout=30) as response: | |
| result = json.loads(response.read().decode("utf-8")) | |
| content = result.get("response", "").strip() | |
| elif interpreter.provider.value == "fake": | |
| # Return fake suggestions for testing | |
| return [ | |
| f"install {user_input} with more details", | |
| f"set up {user_input} environment", | |
| f"configure {user_input} tools", | |
| ] | |
| else: | |
| return [] |
🤖 Prompt for AI Agents
In @cortex/cli.py around lines 130 - 179, The provider comparisons in the chat
branch use interpreter.provider.name (which yields enum member names like
"OPENAI") causing the wrong path to be taken; switch comparisons to use
interpreter.provider.value or compare against the APIProvider enum members
directly (e.g., APIProvider.OPENAI) and update all occurrences in the function
handling the provider logic (the blocks referencing interpreter.provider.name,
interpreter.client, interpreter.model, and interpreter.ollama_url) so each
provider-specific branch (OpenAI/Claude/Ollama/Fake) is matched correctly.
| except Exception: | ||
| # True failure → unknown intent | ||
| return { | ||
| "action": "unknown", | ||
| "domain": "unknown", | ||
| "description": "Failed to extract intent", | ||
| "ambiguous": True, | ||
| "confidence": 0.0, | ||
| } |
There was a problem hiding this comment.
Inconsistent fallback intent structure - missing install_mode field.
The Ollama fallback intent (lines 196-202) is missing the install_mode field that's present in the OpenAI and Claude fallbacks (lines 276-283, 298-305). This could cause KeyError when downstream code accesses intent.get("install_mode").
Add missing field
except Exception:
# True failure → unknown intent
return {
"action": "unknown",
"domain": "unknown",
"description": "Failed to extract intent",
"ambiguous": True,
"confidence": 0.0,
+ "install_mode": "system",
}🤖 Prompt for AI Agents
In @cortex/llm/interpreter.py around lines 194 - 202, The Ollama fallback
returned in the except Exception block in cortex/llm/interpreter.py is missing
the install_mode key, which causes downstream KeyError when code reads
intent.get("install_mode"); update that returned dict (the one currently
returning action/domain/description/ambiguous/confidence) to include
"install_mode": "unknown" (matching the shape used by the OpenAI/Claude
fallbacks) so the intent structure is consistent.
| Format: | ||
| { | ||
| "action": "...", | ||
| "domain": "...", | ||
| "install_mode": "..." | ||
| "description": "...", | ||
| "ambiguous": true/false, | ||
| "confidence": 0.0 | ||
| } | ||
| """ |
There was a problem hiding this comment.
Fix JSON syntax error in the prompt example.
The example JSON in the prompt is missing a comma after "install_mode": "..." on line 237, which could confuse the LLM about the expected output format.
Fix
Format:
{
"action": "...",
"domain": "...",
- "install_mode": "..."
+ "install_mode": "...",
"description": "...",
"ambiguous": true/false,
"confidence": 0.0
}📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| Format: | |
| { | |
| "action": "...", | |
| "domain": "...", | |
| "install_mode": "..." | |
| "description": "...", | |
| "ambiguous": true/false, | |
| "confidence": 0.0 | |
| } | |
| """ | |
| Format: | |
| { | |
| "action": "...", | |
| "domain": "...", | |
| "install_mode": "...", | |
| "description": "...", | |
| "ambiguous": true/false, | |
| "confidence": 0.0 | |
| } | |
| """ |
🤖 Prompt for AI Agents
In @cortex/llm/interpreter.py around lines 233 - 242, The JSON example in the
prompt block under the Format: docstring in cortex/llm/interpreter.py has a
missing comma after the "install_mode": "..." entry; update that prompt/example
(the triple-quoted string starting with "Format:") to include a comma after
"install_mode": "..." so the example becomes valid JSON and won't confuse the
LLM.
| # If parsing fails, do NOT guess meaning | ||
| return { | ||
| "action": "unknown", | ||
| "domain": "unknown", | ||
| "description": "Unstructured intent output", | ||
| "ambiguous": True, | ||
| "confidence": 0.0, | ||
| } |
There was a problem hiding this comment.
Fallback intent missing install_mode field.
The fallback intent returned when parsing fails is also missing install_mode, which should be consistent with other fallback structures.
Fix
# If parsing fails, do NOT guess meaning
return {
"action": "unknown",
"domain": "unknown",
"description": "Unstructured intent output",
"ambiguous": True,
"confidence": 0.0,
+ "install_mode": "system",
}📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| # If parsing fails, do NOT guess meaning | |
| return { | |
| "action": "unknown", | |
| "domain": "unknown", | |
| "description": "Unstructured intent output", | |
| "ambiguous": True, | |
| "confidence": 0.0, | |
| } | |
| # If parsing fails, do NOT guess meaning | |
| return { | |
| "action": "unknown", | |
| "domain": "unknown", | |
| "description": "Unstructured intent output", | |
| "ambiguous": True, | |
| "confidence": 0.0, | |
| "install_mode": "system", | |
| } |
🤖 Prompt for AI Agents
In @cortex/llm/interpreter.py around lines 328 - 335, The fallback intent
returned in the parsing-failure branch is missing the install_mode field; update
the returned dict in the failure branch (the anonymous intent fallback returned
in interpreter.py) to include "install_mode" with the same sentinel value used
by other fallbacks (e.g., "unknown" or the project’s standard fallback value) so
the structure matches other intent fallback objects.
| def extract_intent(self, user_input: str) -> dict: | ||
| if not user_input or not user_input.strip(): | ||
| raise ValueError("User input cannot be empty") | ||
|
|
||
| if self.provider == APIProvider.OPENAI: | ||
| return self._extract_intent_openai(user_input) | ||
| elif self.provider == APIProvider.CLAUDE: | ||
| return self._extract_intent_claude(user_input) | ||
| elif self.provider == APIProvider.OLLAMA: | ||
| return self._extract_intent_ollama(user_input) | ||
| elif self.provider == APIProvider.FAKE: | ||
| # Check for configurable fake intent from environment | ||
| fake_intent_env = os.environ.get("CORTEX_FAKE_INTENT") | ||
| if fake_intent_env: | ||
| try: | ||
| return json.loads(fake_intent_env) | ||
| except json.JSONDecodeError: | ||
| pass # Fall back to default | ||
|
|
||
| # Return realistic intent for testing (not ambiguous) | ||
| return { | ||
| "action": "install", | ||
| "domain": "general", | ||
| "install_mode": "system", | ||
| "description": user_input, | ||
| "ambiguous": False, | ||
| "confidence": 0.8, | ||
| } | ||
| else: | ||
| raise ValueError(f"Unsupported provider: {self.provider}") |
There was a problem hiding this comment.
🛠️ Refactor suggestion | 🟠 Major
Add docstring for public API method.
Per coding guidelines, docstrings are required for all public APIs. The extract_intent method is a public API that should document its purpose, parameters, and return value.
Add docstring
def extract_intent(self, user_input: str) -> dict:
+ """Extract intent from natural language input.
+
+ Args:
+ user_input: Natural language description of desired action
+
+ Returns:
+ Dict containing action, domain, install_mode, description,
+ ambiguous flag, and confidence score
+
+ Raises:
+ ValueError: If input is empty
+ """
if not user_input or not user_input.strip():
raise ValueError("User input cannot be empty")📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| def extract_intent(self, user_input: str) -> dict: | |
| if not user_input or not user_input.strip(): | |
| raise ValueError("User input cannot be empty") | |
| if self.provider == APIProvider.OPENAI: | |
| return self._extract_intent_openai(user_input) | |
| elif self.provider == APIProvider.CLAUDE: | |
| return self._extract_intent_claude(user_input) | |
| elif self.provider == APIProvider.OLLAMA: | |
| return self._extract_intent_ollama(user_input) | |
| elif self.provider == APIProvider.FAKE: | |
| # Check for configurable fake intent from environment | |
| fake_intent_env = os.environ.get("CORTEX_FAKE_INTENT") | |
| if fake_intent_env: | |
| try: | |
| return json.loads(fake_intent_env) | |
| except json.JSONDecodeError: | |
| pass # Fall back to default | |
| # Return realistic intent for testing (not ambiguous) | |
| return { | |
| "action": "install", | |
| "domain": "general", | |
| "install_mode": "system", | |
| "description": user_input, | |
| "ambiguous": False, | |
| "confidence": 0.8, | |
| } | |
| else: | |
| raise ValueError(f"Unsupported provider: {self.provider}") | |
| def extract_intent(self, user_input: str) -> dict: | |
| """Extract intent from natural language input. | |
| Args: | |
| user_input: Natural language description of desired action | |
| Returns: | |
| Dict containing action, domain, install_mode, description, | |
| ambiguous flag, and confidence score | |
| Raises: | |
| ValueError: If input is empty | |
| """ | |
| if not user_input or not user_input.strip(): | |
| raise ValueError("User input cannot be empty") | |
| if self.provider == APIProvider.OPENAI: | |
| return self._extract_intent_openai(user_input) | |
| elif self.provider == APIProvider.CLAUDE: | |
| return self._extract_intent_claude(user_input) | |
| elif self.provider == APIProvider.OLLAMA: | |
| return self._extract_intent_ollama(user_input) | |
| elif self.provider == APIProvider.FAKE: | |
| # Check for configurable fake intent from environment | |
| fake_intent_env = os.environ.get("CORTEX_FAKE_INTENT") | |
| if fake_intent_env: | |
| try: | |
| return json.loads(fake_intent_env) | |
| except json.JSONDecodeError: | |
| pass # Fall back to default | |
| # Return realistic intent for testing (not ambiguous) | |
| return { | |
| "action": "install", | |
| "domain": "general", | |
| "install_mode": "system", | |
| "description": user_input, | |
| "ambiguous": False, | |
| "confidence": 0.8, | |
| } | |
| else: | |
| raise ValueError(f"Unsupported provider: {self.provider}") |
🤖 Prompt for AI Agents
In @cortex/llm/interpreter.py around lines 575 - 604, Add a proper docstring to
the public method extract_intent describing its purpose (parse user input into
an intent dict), the parameters (user_input: str), the return type (dict with
keys like action, domain, install_mode, description, ambiguous, confidence and
that it may vary by provider), and raised exceptions (ValueError for empty input
or unsupported provider, JSONDecodeError handling for CORTEX_FAKE_INTENT).
Mention supported providers (APIProvider.OPENAI, APIProvider.CLAUDE,
APIProvider.OLLAMA, APIProvider.FAKE) and note that when using FAKE it may read
CORTEX_FAKE_INTENT from the environment as JSON fallback behavior.
| # Natural Language Installer (NL Installer) | ||
|
|
||
| Cortex supports installing software using natural language instead of | ||
| explicit package names. | ||
|
|
||
| Example: | ||
| ```bash | ||
| cortex install "something for machine learning" | ||
| ``` | ||
| The request is converted into shell commands using the CommandInterpreter | ||
| By default, commands are generated and printed (dry-run). | ||
| Execution only happens when `--execute` is explicitly provided. | ||
|
|
||
| ```bash | ||
| cortex install "something for machine learning" --execute | ||
| ``` | ||
|
|
||
| The NL installer is validated using unit tests in `tests/test_nl_installer.py`. No newline at end of file |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Check if the referenced test files exist
echo "=== Checking for test_nl_installer.py ==="
fd -t f "test_nl_installer.py" tests/ 2>/dev/null || echo "Not found"
echo "=== Checking for test_nl_parser files ==="
fd -t f "test_nl_parser" tests/ 2>/dev/null || echo "Not found"
echo "=== Listing all test files for context ==="
fd -t f "test_" tests/ | head -20Repository: cortexlinux/cortex
Length of output: 859
🏁 Script executed:
#!/bin/bash
# Get complete list of test files (remove head limit)
echo "=== All test files in tests directory ==="
fd -t f "test_" tests/ | grep -i "nl\|parser\|install" || echo "No matches found"
# Check if there are any files referencing NL installer
echo ""
echo "=== Files referencing NL installer or parser ==="
rg -l "nl.*installer|NL.*Installer" --type md --type py 2>/dev/null | head -10 || echo "No matches"Repository: cortexlinux/cortex
Length of output: 372
Update incorrect test file reference.
Line 17 references tests/test_nl_installer.py, but this file does not exist. The correct test file is tests/test_nl_parser_cases.py. Update the reference accordingly.
🤖 Prompt for AI Agents
In @docs/docs/nl-installer.md around lines 1 - 18, Update the incorrect test
file reference in the NL Installer docs: change the mention of
tests/test_nl_installer.py to the correct tests/test_nl_parser_cases.py in the
docs/docs/nl-installer.md content (the paragraph that currently states "The NL
installer is validated using unit tests in `tests/test_nl_installer.py`").
Ensure the updated line reads that the NL installer is validated using unit
tests in `tests/test_nl_parser_cases.py`.
2 participants
Related Issue
Closes #
Summary
AI Disclosure
Checklist
pytest tests/)Summary by CodeRabbit
Release Notes
New Features
Improvements
Documentation
Removals
✏️ Tip: You can customize this high-level summary in your review settings.