k-cli-for-devs 1.0.3__tar.gz → 1.0.5__tar.gz

{k_cli_for_devs-1.0.3/k_cli_for_devs.egg-info → k_cli_for_devs-1.0.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: k-cli-for-devs
-Version: 1.0.3
+Version: 1.0.5
 Summary: K-CLI for Devs: Autonomous Self-Healing DevOps & Engineering Agent built with AWS Strands Agents SDK
 License: MIT License
         
@@ -89,14 +89,23 @@ pip install k-cli-for-devs
   - [3. Zero-Trust AST Compiler Verification Guardrails](#3-zero-trust-ast-compiler-verification-guardrails)
   - [4. Sub-Millisecond Adaptive Intent Sensor & Smart Router](#4-sub-millisecond-adaptive-intent-sensor--smart-router)
   - [5. Sovereign Air-Gapped Offline Engine](#5-sovereign-air-gapped-offline-engine)
+  - [6. Smart Credit Saver ($2 vs $10 Financial Optimization Engine)](#6-smart-credit-saver-2-vs-10-financial-optimization-engine)
+  - [7. RateLimitGuard & Multi-Provider Model Auto-Rotation](#7-ratelimitguard--multi-provider-model-auto-rotation)
+  - [8. Autonomous Multi-Agent Workstation & Subagent Delegation](#8-autonomous-multi-agent-workstation--subagent-delegation)
+  - [9. Autonomous Time-Travel Checkpoints & Instant Rollback](#9-autonomous-time-travel-checkpoints--instant-rollback-k-cli-undo)
+  - [10. Persistent Self-Learning Project Memory](#10-persistent-self-learning-project-memory-kclimd--projectmemorymanager)
+  - [11. Standardized Quantitative Evaluation & Benchmark Scorecard](#11-standardized-quantitative-evaluation--benchmark-scorecard-k-cli-eval)
+  - [12. Autonomous Docker & CI/CD Pipeline Healer](#12-autonomous-docker--cicd-pipeline-healer-k-cli-cicd)
+  - [13. Global Ambient Error Interceptor Sentinel](#13-global-ambient-error-interceptor-sentinel-k-cli-wrap-cmd)
 - [🖥️ 3 Unified Interface Tiers](#️-3-unified-interface-tiers)
 - [🛡️ The 4 Autonomous Superpowers](#️-the-4-autonomous-superpowers)
-- [🌟 The 13 Production-Grade Killer Features](#-the-13-production-grade-killer-features)
+- [🌟 The 20 Production-Grade Killer Features](#-the-20-production-grade-killer-features)
 - [🏛️ System Architecture Diagram](#️-system-architecture-diagram)
 - [📦 Installation & Quickstart](#-installation--quickstart)
 - [💻 Comprehensive CLI Command Catalog](#-comprehensive-cli-command-catalog)
 - [☁️ Amazon Bedrock AgentCore Export & Cloud Deployment](#️-amazon-bedrock-agentcore-export--cloud-deployment)
 - [🧪 Test Suite & Reliability Scorecard](#-test-suite--reliability-scorecard)
+- [📋 Official Live App Verification Report (16/16 Pass)](docs/LIVE_APP_VERIFICATION_REPORT.md)
 - [🎬 5-Minute Championship Demo Video](#-5-minute-championship-demo-video)
 - [📄 License & Authorship](#-license--authorship)
 
@@ -164,6 +173,9 @@ How does K-CLI compare against leading CLI developer assistants like **Aider**,
 | **100% Air-Gapped Sovereign Offline Mode** | ✅ **Local SLMs (Bankai) + Offline DevDocs** | ⚠️ Partial (Ollama, no offline docs) | ❌ No (Cloud only) | ❌ No (Cloud only) |
 | **Multi-Model Adversarial Consensus** | ✅ **Red Team / Blue Team Swarms** | ❌ No (Single model) | ❌ No (Single model) | ❌ No (Single model) |
 | **Background Auto-Healing Daemon** | ✅ **Autonomous Watcher (`k-cli daemon`)** | ❌ No (Interactive only) | ❌ No (Interactive only) | ❌ No (Cloud webhooks only) |
+| **Smart Credit Saver ($2 vs $10 Benchmark)** | ✅ **Context Pruning & $0.00 AST (85% savings)** | ❌ No (Sends raw context) | ❌ No (Standard pricing) | ❌ No (No optimization) |
+| **RateLimitGuard & Model Auto-Rotation** | ✅ **Circuit Breaker (429 Zero-Downtime)** | ⚠️ Manual fallback | ❌ No (Halts on 429) | ❌ No (Halts on error) |
+| **Autonomous Subagent Spawning** | ✅ **Native Multi-Agent Delegation** | ❌ No (Single agent only) | ⚠️ Partial preview | ❌ No |
 
 ---
 
@@ -204,6 +216,65 @@ For security-sensitive, defense, or offline flight environments, K-CLI operates
 * Includes an embedded SQLite full-text documentation database (DevDocs offline index) covering Python, FastAPI, Docker, Git, and Rust.
 * Transmits **zero telemetry, zero metrics, and zero data leakage**.
 
+### 6. Smart Credit Saver ($2 vs $10 Financial Optimization Engine)
+Most agentic developer tools rapidly burn $10+ of cloud API credits per session by blindly streaming uncompressed 500-line test outputs, massive directory listings, and repetitive AST context into expensive frontier LLMs.
+
+K-CLI’s **Smart Credit Saver** ([`k_cli/core/credit_saver.py`](k_cli/core/credit_saver.py)) slashes token expenditure by **70% to 90%**, enabling complex engineering tasks to execute for **~$1 to $2 instead of $10+**:
+* **Dynamic Context & Output Pruning**: Automatically filters verbose pytest runs, build logs, and file listings down to culprit lines, stack traces, and failure counters, discarding thousands of redundant tokens.
+* **$0.00 Ground-Truth Local AST Verification**: Validates Python syntax natively using the local CPU compiler (`ast.parse` and `py_compile`), completely eliminating costly LLM syntax verification passes.
+* **Sliding Context Window Compaction**: Intelligently condenses intermediate turns while preserving the initial user prompt and ground-truth tool outputs.
+* **Real-Time Financial Telemetry**: Accurately tracks prompt tokens, completion tokens, and dollar savings against an uncompressed frontier baseline, surfaced across the TUI, Web UI, and REST API (e.g. `💰 CreditSaver: Spent $0.18 vs $1.20 baseline (85% saved, 14,200 tokens pruned)`).
+
+### 7. RateLimitGuard & Multi-Provider Model Auto-Rotation
+API rate limits (HTTP 429) and quota exhaustion are the #1 cause of catastrophic failures in autonomous agent loops. K-CLI introduces **RateLimitGuard** ([`k_cli/core/rate_limit_guard.py`](k_cli/core/rate_limit_guard.py)):
+* **Thread-Safe Circuit Breaker**: Automatically catches `429 Too Many Requests`, `RESOURCE_EXHAUSTED`, `RateLimitError`, and `503 Overloaded` exceptions.
+* **Adaptive Jittered Cooldowns**: Enforces an exponential backoff cooldown per provider, preventing repetitive API key throttling.
+* **Zero-Downtime Multi-Provider Auto-Rotation**: When a provider enters cooldown, K-CLI seamlessly and transparently pivots to the next available tier provider with active credentials:
+  * **Fast / Tool Execution**: `Gemini 2.5 Flash` ➔ `Claude 3.5 Haiku` ➔ `GPT-4o-mini` ➔ `DeepSeek Chat` ➔ `Groq` ➔ `Local Ollama / Bankai`.
+  * **Heavy / Architectural Coding**: `Claude 3.7 Sonnet` ➔ `Gemini 2.5 Pro` ➔ `GPT-4o` ➔ `DeepSeek Coder` ➔ `Bankai-14B` ➔ `Deterministic Fallback`.
+* **Zero User Interruption**: The user never experiences a broken build or stopped workflow due to quota or rate-limit spikes.
+
+### 8. Autonomous Multi-Agent Workstation & Subagent Delegation
+Backed by Google Antigravity & Claude Code architectural principles, K-CLI's [`AutonomousAgent`](k_cli/agents/autonomous_agent.py) executes multi-file projects from **just one single command**:
+* **Autonomous Local Tool ReAct Loop**: Invokes `list_dir`, `read_workspace_file`, `write_workspace_file`, `edit_workspace_file`, `execute_command`, `inspect_repo_structure`, `verify_code_file`, `triage_and_heal_incident`, and `heal_cicd_pipeline`.
+* **Specialized Subagent Spawning (`spawn_subagent`)**: Dynamically dispatches background subagents with isolated context windows:
+  * `researcher`: Explores codebase dependencies, symbol maps, and documentation.
+  * `coder`: Synthesizes modular, clean implementations.
+  * `tester`: Generates test suites and executes local ground-truth verification.
+  * `security_auditor`: Audits for OWASP vulnerabilities and secret leaks.
+  * `refactorer`: Performs surgical search/replace refactors.
+  * `explorer`: Inspects directory trees and module boundaries.
+
+### 9. Autonomous Time-Travel Checkpoints & Instant Rollback (`k-cli undo`)
+AI coding assistants can make unintended file edits. K-CLI protects developer repositories with **Zero-Risk Autonomous Checkpointing** ([`k_cli/git/checkpoint.py`](k_cli/git/checkpoint.py)):
+* **Pre-Execution Snapshot**: Automatically snapshots modified code files to `.kcli/checkpoints/` before running any autonomous tool or command.
+* **Instant Rollback (`k-cli undo` / `rollback`)**: One-command restoration reverting modified files to their exact pre-agent state.
+* **Non-Destructive Safety**: Never runs destructive `git reset --hard` or alters uncommitted working branches.
+* **Unified Visual Diff (`k-cli diff-last`)**: Computes exact syntax-highlighted diffs between the active workspace and the last safe checkpoint.
+
+### 10. Persistent Self-Learning Project Memory (`KCLI.md` & `ProjectMemoryManager`)
+Most agents forget past architectural rules across sessions. K-CLI implements **Self-Learning Project Memory** ([`k_cli/core/memory.py`](k_cli/core/memory.py)):
+* **Repository-Specific Context (`KCLI.md`)**: Automatically tracks tech stack standards, verification commands, and forbidden patterns.
+* **Autonomous Feedback Ingestion**: Whenever a bug is healed or a test is fixed, K-CLI records the lesson into `KCLI.md` so the mistake is never repeated.
+* **Bounded Prompt Injection**: Automatically injects high-signal project memory directly into the autonomous agent's system prompt while strictly bounding tokens to avoid context bloat.
+
+### 11. Standardized Quantitative Evaluation & Benchmark Scorecard (`k-cli eval`)
+To prove production reliability, K-CLI features a **Standardized 5-Battery Evaluation Harness** ([`k_cli/tools/benchmark_harness.py`](k_cli/tools/benchmark_harness.py)):
+* **100% Ground-Truth AST Verification**: Evaluates compiler syntax healing, multi-language crash triage, AST security scanning, 3-way merge conflicts, and ReAct agent synthesis.
+* **Real-World Cost Audit**: Measures exact token and financial savings ($ spent vs. unoptimized $10 baseline).
+* **Official Markdown Scorecard**: Exports `.kcli/BENCHMARK_SCORECARD.md` with pass rates and runtime latency metrics for hackathon validation.
+
+### 12. Autonomous Docker & CI/CD Pipeline Healer (`k-cli cicd`)
+Broken CI/CD workflows and bloated container layers are automatically diagnosed and resolved ([`k_cli/tools/cicd_healer.py`](k_cli/tools/cicd_healer.py)):
+* **GitHub Actions Modernization**: Upgrades legacy action versions (`checkout@v2/v3` ➔ `v4`, `setup-python@v3/v4` ➔ `v5`, Node 20 runtime) and injects `PYTHONPATH=.` for seamless test execution.
+* **Dockerfile Optimization**: Injects `--no-cache` to Alpine `apk add`, cleans `/var/lib/apt/lists/*` on Debian/Ubuntu, and adds `--no-cache-dir` to container `pip` calls.
+
+### 13. Global Ambient Error Interceptor Sentinel (`k-cli wrap <cmd>`)
+K-CLI's ambient copilot intercepts terminal errors the microsecond they happen ([`k_cli/tools/sentinel.py`](k_cli/tools/sentinel.py)):
+* **Sub-Second Interception (< 0.1s)**: Intercepts shell, pip, python runtime exceptions, and git merge conflict failures.
+* **Instant Auto-Remediation**: Automatically resolves missing dependencies, `--break-system-packages` restrictions, missing Python interpreter paths, and syntax crashes.
+* **Transparent Re-Execution**: Auto-retries the command upon applying the verified fix, ensuring uninterrupted developer flow.
+
 ---
 
 ## 🖥️ 3 Unified Interface Tiers
@@ -248,7 +319,7 @@ K-CLI provides 3 purpose-built interfaces tailored to developer workflows:
 
 ---
 
-## 🌟 The 13 Production-Grade Killer Features
+## 🌟 The 20 Production-Grade Killer Features
 
 1. **`k-cli exec` / `cmd`**: Google Antigravity-style host machine shell command execution with environment auto-injection.
 2. **`k-cli watch`**: Autonomous PR Review & Watcher Daemon that monitors repos, reviews diffs, and auto-merges clean PRs.
@@ -263,6 +334,13 @@ K-CLI provides 3 purpose-built interfaces tailored to developer workflows:
 11. **`k-cli scaffold`**: Natural Language Full-Stack Scaffolder building complete microservices and APIs from a single prompt.
 12. **`k-cli strands`**: AWS Strands Autonomous Developer Agent executing multi-step engineering goals.
 13. **`k-cli immune`**: Autonomous Chaos Immunity Engine probing brittle AST nodes and synthesizing edge-case tests.
+14. **Smart Credit Saver (`CreditSaver`)**: Dynamic context compaction and zero-cost local AST verification slashing API costs by 70-90% ($1-2 vs $10+).
+15. **RateLimitGuard & Model Auto-Rotator**: Zero-downtime multi-provider circuit breaker auto-rotating across Gemini, Claude, OpenAI, DeepSeek, and Ollama on HTTP 429 rate limits.
+16. **Time-Travel Checkpoints (`k-cli undo` / `checkpoints` / `diff-last`)**: Non-destructive workspace snapshots with instant 1-command rollback.
+17. **Self-Learning Project Memory (`k-cli memory`)**: Persistent `KCLI.md` memory remembering architectural rules and past bug solutions.
+18. **Standardized Evaluation Benchmark (`k-cli eval` / `benchmark`)**: 5-battery quantitative evaluation measuring 100% AST pass rate and financial savings scorecard.
+19. **Docker & CI/CD Pipeline Healer (`k-cli cicd`)**: Automated repair of broken GitHub Actions workflows and unoptimized Dockerfiles.
+20. **Global Ambient Error Interceptor (`k-cli wrap <cmd>` / `sentinel`)**: Sub-second error interceptor fixing pip, python runtime, and git failures on the fly.
 
 ---
 

{k_cli_for_devs-1.0.3 → k_cli_for_devs-1.0.5}/README.md

@@ -33,14 +33,23 @@ pip install k-cli-for-devs
   - [3. Zero-Trust AST Compiler Verification Guardrails](#3-zero-trust-ast-compiler-verification-guardrails)
   - [4. Sub-Millisecond Adaptive Intent Sensor & Smart Router](#4-sub-millisecond-adaptive-intent-sensor--smart-router)
   - [5. Sovereign Air-Gapped Offline Engine](#5-sovereign-air-gapped-offline-engine)
+  - [6. Smart Credit Saver ($2 vs $10 Financial Optimization Engine)](#6-smart-credit-saver-2-vs-10-financial-optimization-engine)
+  - [7. RateLimitGuard & Multi-Provider Model Auto-Rotation](#7-ratelimitguard--multi-provider-model-auto-rotation)
+  - [8. Autonomous Multi-Agent Workstation & Subagent Delegation](#8-autonomous-multi-agent-workstation--subagent-delegation)
+  - [9. Autonomous Time-Travel Checkpoints & Instant Rollback](#9-autonomous-time-travel-checkpoints--instant-rollback-k-cli-undo)
+  - [10. Persistent Self-Learning Project Memory](#10-persistent-self-learning-project-memory-kclimd--projectmemorymanager)
+  - [11. Standardized Quantitative Evaluation & Benchmark Scorecard](#11-standardized-quantitative-evaluation--benchmark-scorecard-k-cli-eval)
+  - [12. Autonomous Docker & CI/CD Pipeline Healer](#12-autonomous-docker--cicd-pipeline-healer-k-cli-cicd)
+  - [13. Global Ambient Error Interceptor Sentinel](#13-global-ambient-error-interceptor-sentinel-k-cli-wrap-cmd)
 - [🖥️ 3 Unified Interface Tiers](#️-3-unified-interface-tiers)
 - [🛡️ The 4 Autonomous Superpowers](#️-the-4-autonomous-superpowers)
-- [🌟 The 13 Production-Grade Killer Features](#-the-13-production-grade-killer-features)
+- [🌟 The 20 Production-Grade Killer Features](#-the-20-production-grade-killer-features)
 - [🏛️ System Architecture Diagram](#️-system-architecture-diagram)
 - [📦 Installation & Quickstart](#-installation--quickstart)
 - [💻 Comprehensive CLI Command Catalog](#-comprehensive-cli-command-catalog)
 - [☁️ Amazon Bedrock AgentCore Export & Cloud Deployment](#️-amazon-bedrock-agentcore-export--cloud-deployment)
 - [🧪 Test Suite & Reliability Scorecard](#-test-suite--reliability-scorecard)
+- [📋 Official Live App Verification Report (16/16 Pass)](docs/LIVE_APP_VERIFICATION_REPORT.md)
 - [🎬 5-Minute Championship Demo Video](#-5-minute-championship-demo-video)
 - [📄 License & Authorship](#-license--authorship)
 
@@ -108,6 +117,9 @@ How does K-CLI compare against leading CLI developer assistants like **Aider**,
 | **100% Air-Gapped Sovereign Offline Mode** | ✅ **Local SLMs (Bankai) + Offline DevDocs** | ⚠️ Partial (Ollama, no offline docs) | ❌ No (Cloud only) | ❌ No (Cloud only) |
 | **Multi-Model Adversarial Consensus** | ✅ **Red Team / Blue Team Swarms** | ❌ No (Single model) | ❌ No (Single model) | ❌ No (Single model) |
 | **Background Auto-Healing Daemon** | ✅ **Autonomous Watcher (`k-cli daemon`)** | ❌ No (Interactive only) | ❌ No (Interactive only) | ❌ No (Cloud webhooks only) |
+| **Smart Credit Saver ($2 vs $10 Benchmark)** | ✅ **Context Pruning & $0.00 AST (85% savings)** | ❌ No (Sends raw context) | ❌ No (Standard pricing) | ❌ No (No optimization) |
+| **RateLimitGuard & Model Auto-Rotation** | ✅ **Circuit Breaker (429 Zero-Downtime)** | ⚠️ Manual fallback | ❌ No (Halts on 429) | ❌ No (Halts on error) |
+| **Autonomous Subagent Spawning** | ✅ **Native Multi-Agent Delegation** | ❌ No (Single agent only) | ⚠️ Partial preview | ❌ No |
 
 ---
 
@@ -148,6 +160,65 @@ For security-sensitive, defense, or offline flight environments, K-CLI operates
 * Includes an embedded SQLite full-text documentation database (DevDocs offline index) covering Python, FastAPI, Docker, Git, and Rust.
 * Transmits **zero telemetry, zero metrics, and zero data leakage**.
 
+### 6. Smart Credit Saver ($2 vs $10 Financial Optimization Engine)
+Most agentic developer tools rapidly burn $10+ of cloud API credits per session by blindly streaming uncompressed 500-line test outputs, massive directory listings, and repetitive AST context into expensive frontier LLMs.
+
+K-CLI’s **Smart Credit Saver** ([`k_cli/core/credit_saver.py`](k_cli/core/credit_saver.py)) slashes token expenditure by **70% to 90%**, enabling complex engineering tasks to execute for **~$1 to $2 instead of $10+**:
+* **Dynamic Context & Output Pruning**: Automatically filters verbose pytest runs, build logs, and file listings down to culprit lines, stack traces, and failure counters, discarding thousands of redundant tokens.
+* **$0.00 Ground-Truth Local AST Verification**: Validates Python syntax natively using the local CPU compiler (`ast.parse` and `py_compile`), completely eliminating costly LLM syntax verification passes.
+* **Sliding Context Window Compaction**: Intelligently condenses intermediate turns while preserving the initial user prompt and ground-truth tool outputs.
+* **Real-Time Financial Telemetry**: Accurately tracks prompt tokens, completion tokens, and dollar savings against an uncompressed frontier baseline, surfaced across the TUI, Web UI, and REST API (e.g. `💰 CreditSaver: Spent $0.18 vs $1.20 baseline (85% saved, 14,200 tokens pruned)`).
+
+### 7. RateLimitGuard & Multi-Provider Model Auto-Rotation
+API rate limits (HTTP 429) and quota exhaustion are the #1 cause of catastrophic failures in autonomous agent loops. K-CLI introduces **RateLimitGuard** ([`k_cli/core/rate_limit_guard.py`](k_cli/core/rate_limit_guard.py)):
+* **Thread-Safe Circuit Breaker**: Automatically catches `429 Too Many Requests`, `RESOURCE_EXHAUSTED`, `RateLimitError`, and `503 Overloaded` exceptions.
+* **Adaptive Jittered Cooldowns**: Enforces an exponential backoff cooldown per provider, preventing repetitive API key throttling.
+* **Zero-Downtime Multi-Provider Auto-Rotation**: When a provider enters cooldown, K-CLI seamlessly and transparently pivots to the next available tier provider with active credentials:
+  * **Fast / Tool Execution**: `Gemini 2.5 Flash` ➔ `Claude 3.5 Haiku` ➔ `GPT-4o-mini` ➔ `DeepSeek Chat` ➔ `Groq` ➔ `Local Ollama / Bankai`.
+  * **Heavy / Architectural Coding**: `Claude 3.7 Sonnet` ➔ `Gemini 2.5 Pro` ➔ `GPT-4o` ➔ `DeepSeek Coder` ➔ `Bankai-14B` ➔ `Deterministic Fallback`.
+* **Zero User Interruption**: The user never experiences a broken build or stopped workflow due to quota or rate-limit spikes.
+
+### 8. Autonomous Multi-Agent Workstation & Subagent Delegation
+Backed by Google Antigravity & Claude Code architectural principles, K-CLI's [`AutonomousAgent`](k_cli/agents/autonomous_agent.py) executes multi-file projects from **just one single command**:
+* **Autonomous Local Tool ReAct Loop**: Invokes `list_dir`, `read_workspace_file`, `write_workspace_file`, `edit_workspace_file`, `execute_command`, `inspect_repo_structure`, `verify_code_file`, `triage_and_heal_incident`, and `heal_cicd_pipeline`.
+* **Specialized Subagent Spawning (`spawn_subagent`)**: Dynamically dispatches background subagents with isolated context windows:
+  * `researcher`: Explores codebase dependencies, symbol maps, and documentation.
+  * `coder`: Synthesizes modular, clean implementations.
+  * `tester`: Generates test suites and executes local ground-truth verification.
+  * `security_auditor`: Audits for OWASP vulnerabilities and secret leaks.
+  * `refactorer`: Performs surgical search/replace refactors.
+  * `explorer`: Inspects directory trees and module boundaries.
+
+### 9. Autonomous Time-Travel Checkpoints & Instant Rollback (`k-cli undo`)
+AI coding assistants can make unintended file edits. K-CLI protects developer repositories with **Zero-Risk Autonomous Checkpointing** ([`k_cli/git/checkpoint.py`](k_cli/git/checkpoint.py)):
+* **Pre-Execution Snapshot**: Automatically snapshots modified code files to `.kcli/checkpoints/` before running any autonomous tool or command.
+* **Instant Rollback (`k-cli undo` / `rollback`)**: One-command restoration reverting modified files to their exact pre-agent state.
+* **Non-Destructive Safety**: Never runs destructive `git reset --hard` or alters uncommitted working branches.
+* **Unified Visual Diff (`k-cli diff-last`)**: Computes exact syntax-highlighted diffs between the active workspace and the last safe checkpoint.
+
+### 10. Persistent Self-Learning Project Memory (`KCLI.md` & `ProjectMemoryManager`)
+Most agents forget past architectural rules across sessions. K-CLI implements **Self-Learning Project Memory** ([`k_cli/core/memory.py`](k_cli/core/memory.py)):
+* **Repository-Specific Context (`KCLI.md`)**: Automatically tracks tech stack standards, verification commands, and forbidden patterns.
+* **Autonomous Feedback Ingestion**: Whenever a bug is healed or a test is fixed, K-CLI records the lesson into `KCLI.md` so the mistake is never repeated.
+* **Bounded Prompt Injection**: Automatically injects high-signal project memory directly into the autonomous agent's system prompt while strictly bounding tokens to avoid context bloat.
+
+### 11. Standardized Quantitative Evaluation & Benchmark Scorecard (`k-cli eval`)
+To prove production reliability, K-CLI features a **Standardized 5-Battery Evaluation Harness** ([`k_cli/tools/benchmark_harness.py`](k_cli/tools/benchmark_harness.py)):
+* **100% Ground-Truth AST Verification**: Evaluates compiler syntax healing, multi-language crash triage, AST security scanning, 3-way merge conflicts, and ReAct agent synthesis.
+* **Real-World Cost Audit**: Measures exact token and financial savings ($ spent vs. unoptimized $10 baseline).
+* **Official Markdown Scorecard**: Exports `.kcli/BENCHMARK_SCORECARD.md` with pass rates and runtime latency metrics for hackathon validation.
+
+### 12. Autonomous Docker & CI/CD Pipeline Healer (`k-cli cicd`)
+Broken CI/CD workflows and bloated container layers are automatically diagnosed and resolved ([`k_cli/tools/cicd_healer.py`](k_cli/tools/cicd_healer.py)):
+* **GitHub Actions Modernization**: Upgrades legacy action versions (`checkout@v2/v3` ➔ `v4`, `setup-python@v3/v4` ➔ `v5`, Node 20 runtime) and injects `PYTHONPATH=.` for seamless test execution.
+* **Dockerfile Optimization**: Injects `--no-cache` to Alpine `apk add`, cleans `/var/lib/apt/lists/*` on Debian/Ubuntu, and adds `--no-cache-dir` to container `pip` calls.
+
+### 13. Global Ambient Error Interceptor Sentinel (`k-cli wrap <cmd>`)
+K-CLI's ambient copilot intercepts terminal errors the microsecond they happen ([`k_cli/tools/sentinel.py`](k_cli/tools/sentinel.py)):
+* **Sub-Second Interception (< 0.1s)**: Intercepts shell, pip, python runtime exceptions, and git merge conflict failures.
+* **Instant Auto-Remediation**: Automatically resolves missing dependencies, `--break-system-packages` restrictions, missing Python interpreter paths, and syntax crashes.
+* **Transparent Re-Execution**: Auto-retries the command upon applying the verified fix, ensuring uninterrupted developer flow.
+
 ---
 
 ## 🖥️ 3 Unified Interface Tiers
@@ -192,7 +263,7 @@ K-CLI provides 3 purpose-built interfaces tailored to developer workflows:
 
 ---
 
-## 🌟 The 13 Production-Grade Killer Features
+## 🌟 The 20 Production-Grade Killer Features
 
 1. **`k-cli exec` / `cmd`**: Google Antigravity-style host machine shell command execution with environment auto-injection.
 2. **`k-cli watch`**: Autonomous PR Review & Watcher Daemon that monitors repos, reviews diffs, and auto-merges clean PRs.
@@ -207,6 +278,13 @@ K-CLI provides 3 purpose-built interfaces tailored to developer workflows:
 11. **`k-cli scaffold`**: Natural Language Full-Stack Scaffolder building complete microservices and APIs from a single prompt.
 12. **`k-cli strands`**: AWS Strands Autonomous Developer Agent executing multi-step engineering goals.
 13. **`k-cli immune`**: Autonomous Chaos Immunity Engine probing brittle AST nodes and synthesizing edge-case tests.
+14. **Smart Credit Saver (`CreditSaver`)**: Dynamic context compaction and zero-cost local AST verification slashing API costs by 70-90% ($1-2 vs $10+).
+15. **RateLimitGuard & Model Auto-Rotator**: Zero-downtime multi-provider circuit breaker auto-rotating across Gemini, Claude, OpenAI, DeepSeek, and Ollama on HTTP 429 rate limits.
+16. **Time-Travel Checkpoints (`k-cli undo` / `checkpoints` / `diff-last`)**: Non-destructive workspace snapshots with instant 1-command rollback.
+17. **Self-Learning Project Memory (`k-cli memory`)**: Persistent `KCLI.md` memory remembering architectural rules and past bug solutions.
+18. **Standardized Evaluation Benchmark (`k-cli eval` / `benchmark`)**: 5-battery quantitative evaluation measuring 100% AST pass rate and financial savings scorecard.
+19. **Docker & CI/CD Pipeline Healer (`k-cli cicd`)**: Automated repair of broken GitHub Actions workflows and unoptimized Dockerfiles.
+20. **Global Ambient Error Interceptor (`k-cli wrap <cmd>` / `sentinel`)**: Sub-second error interceptor fixing pip, python runtime, and git failures on the fly.
 
 ---
 

{k_cli_for_devs-1.0.3 → k_cli_for_devs-1.0.5}/k_cli/agents/autonomous_agent.py

@@ -273,6 +273,19 @@ def tool_spawn_subagent(role: str, task: str) -> str:
         return f"Error executing subagent '{role}': {e}"
 
 
+def tool_heal_cicd_pipeline(file_path: str = ".github/workflows/ci.yml") -> str:
+    """Audits and auto-repairs GitHub Actions workflow or Dockerfile."""
+    try:
+        from k_cli.tools.cicd_healer import global_cicd_healer
+        if "docker" in file_path.lower():
+            res = global_cicd_healer.audit_and_heal_dockerfile(file_path)
+        else:
+            res = global_cicd_healer.audit_and_heal_workflow(file_path)
+        return f"CI/CD Healer for '{file_path}': Success={res.success}, Issues={res.issues_found}, Fixes={res.fixes_applied}"
+    except Exception as e:
+        return f"Error healing CI/CD pipeline: {e}"
+
+
 # Tool registry map
 AVAILABLE_TOOLS: Dict[str, Callable[..., str]] = {
     "list_dir": tool_list_dir,
@@ -285,6 +298,7 @@ AVAILABLE_TOOLS: Dict[str, Callable[..., str]] = {
     "search_codebase": tool_search_codebase,
     "triage_and_heal_incident": tool_triage_and_heal_incident,
     "spawn_subagent": tool_spawn_subagent,
+    "heal_cicd_pipeline": tool_heal_cicd_pipeline,
 }
 
 
@@ -306,12 +320,17 @@ def build_agentic_system_prompt(cwd: Optional[str] = None) -> str:
     
     top_manifest = ", ".join(top_items[:20]) or "standard project tree"
     
+    # Load persistent project memory if available
+    from k_cli.core.memory import global_memory_manager
+    memory_context = global_memory_manager.load_memory(max_chars=3000)
+    memory_section = f"\n\nPERSISTENT PROJECT MEMORY & ARCHITECTURAL CONVENTIONS:\n{memory_context}" if memory_context else ""
+
     return f"""You are K-CLI, a fully autonomous AI DevOps and software engineering workstation running directly on the developer's Linux machine (comparable to Google Antigravity, Aider, and Claude Code).
 
 LOCAL MACHINE ENVIRONMENT:
 - Working Directory: {cwd_path}
 - Workspace Root Items: {top_manifest}
-- Execution Privilege: Full local machine terminal execution & file system read/write via tools.
+- Execution Privilege: Full local machine terminal execution & file system read/write via tools.{memory_section}
 
 CRITICAL OPERATIONAL RULES:
 1. YOU ARE NOT A CHATBOT. You are an autonomous software engineering agent with REAL tools on this local machine.
@@ -342,17 +361,18 @@ Available Tools:
 - `search_codebase(query="...", directory=".")`: Search for symbols or text across files.
 - `triage_and_heal_incident(error_traceback="...")`: Automated crash traceback triage and repair.
 - `spawn_subagent(role="researcher|coder|tester|security_auditor|refactorer|explorer", task="...")`: Delegates subtasks to specialized background subagents (Google Antigravity & Claude Code architecture).
+- `heal_cicd_pipeline(file_path=".github/workflows/ci.yml")`: Audits and repairs broken GitHub Actions workflows and Dockerfiles.
 
 COMMUNICATION & NATURAL LANGUAGE STYLE (MANDATORY):
 - Talk like an elite senior staff engineer (similar to Claude Code, Aider, and Google Antigravity). Direct, natural, authoritative, concise.
 - ZERO ROBOTIC PREAMBLES OR CONVERSATIONAL FILLER:
-  - NEVER output robotic phrases like "Okay, I now have a clear picture of...", "Based on the file structure...", "Based on my analysis...", "Here's why I find it impressive:", "Sure, I'd be happy to help!", "Certainly!", or "In summary...".
-  - Get straight to the technical substance without stating what you are about to do or narrating your cognitive state.
+   - NEVER output robotic phrases like "Okay, I now have a clear picture of...", "Based on the file structure...", "Based on my analysis...", "Here's why I find it impressive:", "Sure, I'd be happy to help!", "Certainly!", or "In summary...".
+   - Get straight to the technical substance without stating what you are about to do or narrating your cognitive state.
 - When reviewing a project or directory:
-  - Deliver a crisp, natural, high-signal technical evaluation.
-  - Highlight key architectural layers, modules, testing infrastructure, and technical design decisions using clean bullet points and precise paths.
+   - Deliver a crisp, natural, high-signal technical evaluation.
+   - Highlight key architectural layers, modules, testing infrastructure, and technical design decisions using clean bullet points and precise paths.
 - When writing or modifying code:
-  - Perform the filesystem and compiler operations silently using tools, then state what was accomplished and verified.
+   - Perform the filesystem and compiler operations silently using tools, then state what was accomplished and verified.
 
 Always take real action. Inspect real files, write real code to disk, and verify before giving your final answer."""
 
@@ -519,6 +539,17 @@ class AutonomousAgent:
         system_prompt = build_agentic_system_prompt(self.cwd)
         steps: List[AgentStep] = []
         tools_executed: List[str] = []
+
+        # Automatic Non-Destructive Time-Travel Checkpoint before autonomous operations
+        from k_cli.git.checkpoint import global_checkpoint_manager
+        try:
+            ckpt_id = global_checkpoint_manager.create_checkpoint(
+                description=f"Autonomous run: {prompt[:80]}"
+            )
+            if token_callback:
+                token_callback("CHECKPOINT", f"🛡️ [Created Time-Travel Checkpoint '{ckpt_id}']\n")
+        except Exception as e:
+            logger.warning(f"Failed to create pre-run checkpoint: {e}")
         
         # 1. Proactive Reconnaissance:
         lower_prompt = prompt.lower()
@@ -634,6 +665,18 @@ class AutonomousAgent:
             completion_tokens=total_completion_tokens,
         )
 
+        if tools_executed:
+            try:
+                from k_cli.core.memory import global_memory_manager
+                tools_str = ", ".join(sorted(set(tools_executed)))
+                clean_task = prompt.strip()[:80]
+                global_memory_manager.record_learning(
+                    lesson=f"Executed task: '{clean_task}' (tools used: {tools_str})",
+                    category="AutonomousTask",
+                )
+            except Exception:
+                pass
+
         return AutonomousAgentResult(
             success=True,
             final_response=final_response,

{k_cli_for_devs-1.0.3 → k_cli_for_devs-1.0.5}/k_cli/cli.py

@@ -17,6 +17,7 @@ import json
 import os
 import shlex
 import sys
+import time
 from pathlib import Path
 
 # Ensure project root is in sys.path for direct CLI script execution
@@ -3241,6 +3242,247 @@ def demo_cmd(
     start_cinematic_demo(speed=speed, act=act)
 
 
+# =============================================================================
+# Feature 1: Autonomous Time-Travel Checkpoints & Instant Rollback (`k-cli undo`)
+# =============================================================================
+@app.command(name="undo", help="Instantly roll back the workspace to the latest safe checkpoint.")
+@app.command(name="rollback", help="Alias for undo: revert to previous checkpoint.")
+def undo_cmd(
+    repo: str = typer.Option(".", "--repo", "-r", help="Repository directory to restore."),
+):
+    """Reverts workspace state to the most recent pre-execution checkpoint."""
+    from k_cli.git.checkpoint import CheckpointManager
+    mgr = CheckpointManager(workspace_dir=repo)
+    success, msg = mgr.rollback_last_checkpoint()
+    if success:
+        console.print(Panel(
+            f"[bold green]✔ Time-Travel Rollback Succeeded[/bold green]\n\n{msg}",
+            title="🛡️ K-CLI Checkpoint Rollback",
+            border_style="green",
+        ))
+    else:
+        console.print(Panel(
+            f"[bold red]✘ Rollback Failed[/bold red]\n\n{msg}",
+            title="🛡️ K-CLI Checkpoint Rollback",
+            border_style="red",
+        ))
+
+
+@app.command(name="checkpoints", help="List all autonomous time-travel snapshots in workspace.")
+def checkpoints_cmd(
+    repo: str = typer.Option(".", "--repo", "-r", help="Repository directory."),
+):
+    """Displays saved workspace checkpoints."""
+    from k_cli.git.checkpoint import CheckpointManager
+    mgr = CheckpointManager(workspace_dir=repo)
+    ckpts = mgr.list_checkpoints()
+    if not ckpts:
+        console.print("[yellow]No checkpoints saved yet in .kcli/checkpoints/[/yellow]")
+        return
+    table = Table(title="🛡️ K-CLI Time-Travel Checkpoints", border_style="cyan")
+    table.add_column("Checkpoint ID", style="cyan bold")
+    table.add_column("Timestamp", style="white")
+    table.add_column("Files Tracked", style="green")
+    table.add_column("Description", style="dim")
+    for c in reversed(ckpts):
+        ts = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(c["timestamp"]))
+        table.add_row(c["checkpoint_id"], ts, str(len(c.get("files_tracked", []))), c.get("description", ""))
+    console.print(table)
+
+
+@app.command(name="diff-last", help="Show unified diff between current workspace and latest checkpoint.")
+def diff_last_cmd(
+    repo: str = typer.Option(".", "--repo", "-r", help="Repository directory."),
+):
+    """Displays diff since last checkpoint."""
+    from k_cli.git.checkpoint import CheckpointManager
+    mgr = CheckpointManager(workspace_dir=repo)
+    diff_text = mgr.compute_diff()
+    if "Zero modifications detected" in diff_text:
+        console.print(f"[bold green]✔ {diff_text}[/bold green]")
+    else:
+        syntax = Syntax(diff_text, "diff", theme="monokai", line_numbers=True)
+        console.print(syntax)
+
+
+# =============================================================================
+# Feature 2: Self-Learning Project Memory (`k-cli memory`)
+# =============================================================================
+@app.command(name="memory", help="View or update self-learning project memory (KCLI.md / .kcli/MEMORY.md).")
+def memory_cmd(
+    action: str = typer.Argument("show", help="Action: 'show', 'init', or 'learn'."),
+    note: Optional[str] = typer.Option(None, "--note", "-n", help="Lesson or directive to record."),
+    repo: str = typer.Option(".", "--repo", "-r", help="Target repository directory."),
+):
+    """Inspects and manages persistent project memory."""
+    from k_cli.core.memory import ProjectMemoryManager
+    mgr = ProjectMemoryManager(workspace_dir=repo)
+    if action == "init":
+        mgr.initialize_if_missing()
+        console.print("[bold green]✔ Initialized persistent KCLI.md project memory.[/bold green]")
+    elif action == "learn" and note:
+        mgr.record_learning(note, category="DeveloperNote")
+        console.print(f"[bold green]✔ Recorded learning:[/bold green] {note}")
+    else:
+        content = mgr.load_memory(max_chars=8000)
+        if not content:
+            mgr.initialize_if_missing()
+            content = mgr.load_memory(max_chars=8000)
+        console.print(Panel(
+            Markdown(content),
+            title="🧠 K-CLI Self-Learning Project Memory (KCLI.md)",
+            border_style="magenta",
+        ))
+
+
+# =============================================================================
+# Feature 3: Standardized Evaluation & Benchmark Scorecard (`k-cli eval` / `benchmark`)
+# =============================================================================
+@app.command(name="eval", help="Run 5-battery standardized benchmark evaluation and export official scorecard.")
+@app.command(name="benchmark", help="Alias for k-cli eval: Run standardized autonomous developer benchmark.")
+def eval_cmd(
+    repo: str = typer.Option(".", "--repo", "-r", help="Repository directory to evaluate."),
+    json_out: bool = typer.Option(False, "--json", help="Output raw JSON benchmark data."),
+):
+    """Executes the quantitative 5-battery benchmark measuring AST pass rate and financial savings."""
+    from k_cli.tools.benchmark_harness import EvaluationHarness
+    harness = EvaluationHarness(workspace_dir=repo)
+    console.print("[bold cyan]⚡ Running K-CLI Autonomous Engineering Benchmark Battery...[/bold cyan]")
+    report = harness.run_full_evaluation()
+
+    if json_out:
+        import dataclasses
+        console.print(json.dumps(dataclasses.asdict(report), indent=2))
+        return
+
+    table = Table(title="🏆 K-CLI Standardized Benchmark Scorecard", border_style="green")
+    table.add_column("Task ID", style="bold cyan")
+    table.add_column("Benchmark Task", style="white")
+    table.add_column("Category", style="yellow")
+    table.add_column("Status", style="bold green")
+    table.add_column("AST Ground-Truth", style="bold green")
+    table.add_column("Time", style="cyan")
+    table.add_column("Cost Spent", style="magenta")
+    table.add_column("Cost Saved", style="green")
+
+    for r in report.results:
+        st = "[bold green]✔ PASS[/bold green]" if r.passed else "[bold red]✘ FAIL[/bold red]"
+        ast_st = "[bold green]✔ VALID[/bold green]" if r.ast_verified else "[bold red]✘ FAIL[/bold red]"
+        table.add_row(
+            r.task_id,
+            r.name,
+            r.category,
+            st,
+            ast_st,
+            f"{r.duration_sec}s",
+            f"${r.actual_cost_usd:.4f}",
+            f"${r.saved_usd:.4f}",
+        )
+    console.print(table)
+
+    summary_panel = Panel(
+        f"[bold bright_white]Overall Pass Rate:[/bold bright_white] [bold green]{report.passed_tasks}/{report.total_tasks} ({report.ast_pass_rate_pct}% AST Verified)[/bold green]\n"
+        f"[bold bright_white]Total Test Duration:[/bold bright_white] [cyan]{report.total_duration_sec}s[/cyan]\n"
+        f"[bold bright_white]Actual Cloud/Model Spend:[/bold bright_white] [magenta]${report.total_spent_usd:.4f}[/magenta]\n"
+        f"[bold bright_white]CreditSaver Optimization:[/bold bright_white] [bold green]${report.total_saved_usd:.4f} saved ({report.savings_pct}% cheaper than $10 unoptimized baseline)[/bold green]\n\n"
+        f"[dim]Scorecard exported to: {repo}/.kcli/BENCHMARK_SCORECARD.md[/dim]",
+        title="📊 Executive Evaluation Metrics",
+        border_style="cyan",
+    )
+    console.print(summary_panel)
+
+
+# =============================================================================
+# Feature 4: Autonomous Docker & CI/CD Pipeline Healer (`k-cli cicd`)
+# =============================================================================
+@app.command(name="cicd", help="Audit and auto-repair broken GitHub Actions workflows and Dockerfiles.")
+def cicd_cmd(
+    target: str = typer.Argument("all", help="Target: 'all', 'workflow', 'dockerfile', or specific path."),
+    repo: str = typer.Option(".", "--repo", "-r", help="Repository directory."),
+    dry_run: bool = typer.Option(False, "--dry-run", help="Audit only without writing changes."),
+):
+    """Diagnoses and heals broken CI/CD pipelines and Dockerfiles."""
+    from k_cli.tools.cicd_healer import CICDHealer
+    healer = CICDHealer(workspace_dir=repo)
+    auto_apply = not dry_run
+
+    console.print(f"[bold cyan]⚡ Running K-CLI CI/CD & Docker Pipeline Healer on '{repo}'...[/bold cyan]\n")
+    results = []
+
+    # Workflow healing
+    wf_path = Path(repo) / ".github" / "workflows"
+    if wf_path.exists():
+        for wf_file in wf_path.glob("*.yml"):
+            results.append(healer.audit_and_heal_workflow(str(wf_file), auto_apply=auto_apply))
+        for wf_file in wf_path.glob("*.yaml"):
+            results.append(healer.audit_and_heal_workflow(str(wf_file), auto_apply=auto_apply))
+
+    # Dockerfile healing
+    df_path = Path(repo) / "Dockerfile"
+    if df_path.exists():
+        results.append(healer.audit_and_heal_dockerfile(str(df_path), auto_apply=auto_apply))
+
+    if not results:
+        console.print("[yellow]No GitHub Actions workflows or Dockerfiles detected in repository.[/yellow]")
+        return
+
+    for res in results:
+        p_name = Path(res.file_path).name
+        if res.issues_found > 0:
+            console.print(Panel(
+                f"[bold green]✔ Healed {res.issues_found} issue(s) in {p_name}:[/bold green]\n" +
+                "\n".join(f"  • {f}" for f in res.fixes_applied),
+                title=f"🔧 CI/CD Healer: {p_name}",
+                border_style="green",
+            ))
+        else:
+            console.print(f"[green]✔[/green] {p_name}: [dim]All action versions, cache directives, and build layers verified optimal.[/dim]")
+
+
+# =============================================================================
+# Feature 5: Global Ambient Error Interceptor Sentinel (`k-cli wrap <cmd>`)
+# =============================================================================
+@app.command(name="wrap", help="Run ANY shell/pip/git command under ambient Sentinel supervision; auto-fixes errors in <1s.")
+@app.command(name="sentinel", help="Alias for k-cli wrap: Ambient zero-latency error interceptor.")
+def wrap_cmd(
+    command: List[str] = typer.Argument(..., help="Shell command to execute and monitor."),
+    repo: str = typer.Option(".", "--repo", "-r", help="Working directory for execution."),
+):
+    """Runs a command with Global Sentinel active. If any error occurs, Sentinel intercepts and heals it instantly."""
+    from k_cli.tools.sentinel import GlobalSentinel
+    import shlex
+    cmd_str = shlex.join(command)
+    console.print(f"[bold cyan]⚡ K-CLI Global Sentinel active on: [bold white]{cmd_str}[/bold white][/bold cyan]\n")
+    sentinel = GlobalSentinel(workspace_dir=repo)
+    result = sentinel.wrap_and_heal(cmd_str, cwd=repo)
+
+    if result.stdout.strip():
+        console.print(result.stdout.strip())
+    if result.stderr.strip() and result.final_exit_code != 0:
+        console.print(f"[red]{result.stderr.strip()}[/red]")
+
+    if result.original_exit_code == 0:
+        console.print(f"\n[bold green]✔ Command completed successfully ({result.duration_sec}s)[/bold green]")
+    elif result.repair_successful:
+        console.print(Panel(
+            f"[bold green]✔ Sentinel Auto-Repaired Command in {result.duration_sec}s[/bold green]\n\n"
+            f"• [cyan]Intercepted Error:[/cyan] {result.culprit_detected}\n"
+            f"• [yellow]Action Taken:[/yellow] {result.repair_action}\n"
+            f"• [green]Final Status:[/green] Exit Code {result.final_exit_code} (VERIFIED RE-EXECUTION SUCCESS)",
+            title="🛡️ K-CLI Global Sentinel Interception",
+            border_style="green",
+        ))
+    else:
+        console.print(Panel(
+            f"[bold red]✘ Sentinel Intercepted Error (Exit Code {result.final_exit_code})[/bold red]\n\n"
+            f"• [cyan]Culprit:[/cyan] {result.culprit_detected}\n"
+            f"• [yellow]Intervention:[/yellow] {result.repair_action}",
+            title="⚠️ K-CLI Sentinel Warning",
+            border_style="yellow",
+        ))
+        raise typer.Exit(code=result.final_exit_code)
+
+
 def interactive_mode(model: str = "qwen2.5-coder:1.5b", mock: bool = False, continue_session: bool = False):
     """Interactive multi-turn prompt shell when typing 'k' without arguments."""
     if hasattr(console, "is_terminal") and console.is_terminal: