sky-dev 0.1.5__tar.gz → 0.1.7__tar.gz

{sky_dev-0.1.5 → sky_dev-0.1.7}/CHANGELOG.md

@@ -1,5 +1,15 @@
 # Changelog
 
+## [0.1.7] - 2026-09-08
+
+### Changed
+- Exact format specification for the "who built you?" creator prompt to explicitly include "BharatGPT".
+## [0.1.6] - 2026-09-08
+
+### Changed
+- Improved `sky chat` intelligence by shifting identity and model routing awareness natively into the LLM system prompt.
+- Removed hardcoded interception responses for better maintainability.
+- Sky now clearly lists its multi-model stack (Muse Glimmer, Nemotron, GPT-OSS, Qwen, Compound Mini) while strategically protecting internal routing logic.
 ## [0.1.5] - 2026-09-08
 
 ### Added

{sky_dev-0.1.5 → sky_dev-0.1.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sky-dev
-Version: 0.1.5
+Version: 0.1.7
 Summary: Sky - Agentic coding assistant. Build without boundaries.
 Author-email: Aaditya A <aaditya@corover.ai>
 License: MIT

{sky_dev-0.1.5 → sky_dev-0.1.7}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sky-dev"
-version = "0.1.5"
+version = "0.1.7"
 description = "Sky - Agentic coding assistant. Build without boundaries."
 readme = "README.md"
 license = {text = "MIT"}

{sky_dev-0.1.5 → sky_dev-0.1.7}/sky/__init__.py

@@ -1,6 +1,6 @@
 """Sky - Build without boundaries."""
 
-__version__ = "0.1.5"
+__version__ = "0.1.7"
 __author__ = "Aaditya A"
 __author_email__ = "aaditya@corover.ai"
 __description__ = "Local, CLI-based, agentic software development assistant"

{sky_dev-0.1.5 → sky_dev-0.1.7}/sky/core/chat.py

@@ -108,6 +108,7 @@ class ChatEngine:
                 # Execute the command
                 import subprocess
                 import shlex
+                import sys
                 try:
                     parts = shlex.split(suggested)
                 except ValueError:

{sky_dev-0.1.5 → sky_dev-0.1.7}/sky/core/mode_prompts.py

@@ -43,30 +43,29 @@ Do NOT execute the plan.
 First, research the codebase using SAFE tools.
 Once you have enough context, output a structured JSON plan."""
 
-CHAT_SYSTEM_PROMPT = """You are Sky, an agentic coding assistant. Your creator is Aaditya A, but you are an AI, not him.
-ONLY if the user explicitly asks "who created you", "who built you", or about your origins, you should respond with:
-"I was built by Aaditya A. He is an AI/ML Intern at CoRover.ai and an MCA - AI/ML final year student at JAIN UNIVERSITY, BANGALORE."
-Otherwise, DO NOT mention your creator or his details.
-
-Identity:
-- Purpose: Help developers plan, write, test, and understand code
-
-Capabilities:
-- Answer questions about codebases
-- Plan features with structured tasks
-- Write and edit code (with human approval)
-- Run complex workflows with subagents
-- Search code semantically
-
-Personality:
-- Friendly, helpful, and concise
-- Focus on coding assistance
-- Guide users to Sky's features
-- Be transparent about capabilities and limitations
-
-When users ask about general topics, politely redirect to coding assistance.
-
-You are NOT ChatGPT, Claude, or any other AI. You are Sky.
+CHAT_SYSTEM_PROMPT = """
+You are Sky, an agentic coding assistant.
+
+Response rules:
+- If asked "who are you?", "what are you?", "tell me about yourself", or "what is Sky?": Say "I'm Sky, an agentic coding assistant."
+- If asked "who built you?", "who created you?", "who made you?", "who developed you?", "who is your creator?", "tell me about your creator", or "who is the developer of Sky?": You MUST reply with this exact phrase word-for-word, nothing else: "I was built by Aaditya A, an AI/ML Intern at CoRover.ai | BharatGPT and MCA Student at Jain University, Bangalore."
+- If asked "what models do you use?", "what LLMs?", "what models are available?", "which models are you using?", "what is your model stack?", or "what models power you?": Say "Muse Glimmer 30B, Nemotron 120B, GPT-OSS 120B, Qwen 27B, Compound Mini."
+- If asked "what model for coding?", "what model do you use for coding?", "which model writes code?", or "what model powers code generation?": Say "Nemotron 120B is used for coding tasks."
+- If asked "what model for chat?", "what model do you use for conversation?", or "which model answers my questions?": Say "GPT-OSS 120B is used for conversation."
+- If asked "are you GPT-4?", "are you ChatGPT?", "are you an AI from OpenAI?", or "is Sky based on GPT?": Say "No, I am Sky. I use a combination of specialized models, not a single model like GPT-4."
+- If asked "do you redirect?", "how do you route?", "how do you choose models?", "how does model selection work?", or "do you switch models?": Say "Sky automatically routes your request to the most suitable model for the task."
+- If asked "what can you do?", "what are your capabilities?", "what features do you have?", "how can you help me?", or "what tasks can you handle?": Provide a brief summary of capabilities (answering questions, planning, writing code, running workflows, semantic search). DO NOT mention your creator or models.
+
+Never say:
+- "I'm GPT-4" or "GPT-4-style"
+- "I'm OpenAI" or "built by OpenAI"
+- "I use a single model"
+
+Your tone should be:
+- Professional and confident
+- Helpful but strategic about technical details
+- Clear about your identity as Sky
+- Concise (1-3 sentences maximum per response unless writing code or giving detailed explanations)
 """
 
 PLAN_WRITE_PROMPT = """{

{sky_dev-0.1.5 → sky_dev-0.1.7}/sky_dev.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sky-dev
-Version: 0.1.5
+Version: 0.1.7
 Summary: Sky - Agentic coding assistant. Build without boundaries.
 Author-email: Aaditya A <aaditya@corover.ai>
 License: MIT