npcsh 1.1.16__py3-none-any.whl → 1.1.18__py3-none-any.whl

npcsh/npc_team/jinxs/bin/benchmark.jinx

@@ -0,0 +1,146 @@
+jinx_name: benchmark
+description: Run Terminal-Bench evaluation to benchmark npcsh performance with different models
+inputs:
+- model: ""
+- provider: ""
+- action: "check"
+- concurrent: "4"
+- npc_name: ""
+
+steps:
+  - name: run_benchmark
+    engine: python
+    code: |
+      import os
+      import sys
+
+      action = {{ action | default("check") | tojson }}.strip().lower()
+      model = {{ model | default("") | tojson }}.strip()
+      provider = {{ provider | default("") | tojson }}.strip()
+      concurrent = int({{ concurrent | default("4") | tojson }} or "4")
+      npc_name_input = {{ npc_name | default("") | tojson }}.strip() or None
+
+      if not model:
+          model = npc.model if npc and npc.model else ""
+      if not provider:
+          provider = npc.provider if npc and npc.provider else "anthropic"
+
+      try:
+          from npcsh.benchmark import BenchmarkRunner, run_benchmark
+          BENCHMARK_AVAILABLE = True
+      except ImportError:
+          BENCHMARK_AVAILABLE = False
+
+      if action == "check":
+          output = "## Terminal-Bench Integration Status\n\n"
+
+          if not BENCHMARK_AVAILABLE:
+              output += "**Status:** Benchmark module not fully loaded (harbor not installed)\n\n"
+          else:
+              output += "**Status:** Ready\n\n"
+
+          if BENCHMARK_AVAILABLE:
+              runner = BenchmarkRunner()
+              deps = runner.check_dependencies()
+              output += "### Dependencies:\n"
+              for dep, installed in deps.items():
+                  status = "Installed" if installed else "Not installed"
+                  output += "- **{}**: {}\n".format(dep, status)
+
+              if not all(deps.values()):
+                  output += "\n### Installation:\n"
+                  output += "```bash\n"
+                  output += "pip install harbor terminal-bench\n"
+                  output += "```\n"
+
+          output += "\n### Usage:\n"
+          output += "```\n"
+          output += "/benchmark action=quick\n"
+          output += "/benchmark action=run model=gpt-4o provider=openai\n"
+          output += "/benchmark action=list\n"
+          output += "```\n"
+
+      elif action == "list":
+          if not BENCHMARK_AVAILABLE:
+              output = "Error: Benchmark module not available. Run `/benchmark` first."
+          else:
+              runner = BenchmarkRunner()
+              runs = runner.list_past_runs()
+
+              if not runs:
+                  output = "No past benchmark runs found."
+              else:
+                  output = "## Past Benchmark Runs ({} total)\n\n".format(len(runs))
+                  for run in runs[:10]:
+                      timestamp = run.get('timestamp', 'unknown')[:19]
+                      model_name = run.get('model', 'unknown')
+                      result = run.get('result', {})
+                      accuracy = result.get('accuracy', 0)
+                      passed = result.get('passed_tasks', 0)
+                      total = result.get('total_tasks', 0)
+
+                      output += "### {}\n".format(timestamp)
+                      output += "- **Model:** {}\n".format(model_name)
+                      output += "- **Accuracy:** {:.1%}\n".format(accuracy)
+                      output += "- **Tasks:** {}/{}\n\n".format(passed, total)
+
+      elif action == "quick":
+          if not BENCHMARK_AVAILABLE:
+              output = "Error: Install with: pip install harbor terminal-bench"
+          else:
+              output = "## Quick Test: {}/{}\n\n".format(provider, model)
+              output += "Running quick test with 3 tasks...\n\n"
+
+              try:
+                  from npcsh.benchmark import quick_test
+                  result = quick_test(model=model, provider=provider)
+
+                  status = "PASS" if result.success else "FAIL"
+                  output += "**Status:** {}\n".format(status)
+                  output += "**Accuracy:** {:.1%}\n".format(result.accuracy)
+                  output += "**Tasks:** {}/{}\n".format(result.passed_tasks, result.total_tasks)
+                  output += "**Duration:** {:.1f}s\n".format(result.duration_seconds)
+
+                  if result.error:
+                      output += "\n**Error:** {}\n".format(result.error)
+
+                  output += "\n**Output:** {}\n".format(result.output_dir)
+
+              except Exception as e:
+                  output = "Error running quick test: {}".format(e)
+
+      elif action == "run":
+          if not BENCHMARK_AVAILABLE:
+              output = "Error: Install with: pip install harbor terminal-bench"
+          else:
+              output = "## Benchmark Run: {}/{}\n\n".format(provider, model)
+              output += "Running Terminal-Bench 2.0 with {} concurrent tasks...\n\n".format(concurrent)
+
+              try:
+                  runner = BenchmarkRunner()
+                  result = runner.run(
+                      model=model,
+                      provider=provider,
+                      n_concurrent=concurrent,
+                      npc_name=npc_name_input,
+                  )
+
+                  status = "SUCCESS" if result.success else "FAILED"
+                  output += "**Status:** {}\n".format(status)
+                  output += "**Accuracy:** {:.1%}\n".format(result.accuracy)
+                  output += "**Tasks Passed:** {}/{}\n".format(result.passed_tasks, result.total_tasks)
+                  output += "**Duration:** {:.1f}s\n".format(result.duration_seconds)
+                  output += "**Total Tokens:** {:,}\n".format(result.total_tokens)
+                  output += "**Total Cost:** ${:.4f}\n".format(result.total_cost_usd)
+
+                  if result.error:
+                      output += "\n**Error:** {}\n".format(result.error)
+
+                  output += "\n**Results saved to:** {}\n".format(result.output_dir)
+
+              except Exception as e:
+                  import traceback
+                  output = "Error running benchmark: {}\n\n{}".format(e, traceback.format_exc())
+
+      else:
+          output = "Unknown action: {}\n\nAvailable: check, run, quick, list".format(action)

npcsh/npc_team/jinxs/bin/nql.jinx

@@ -1,13 +1,13 @@
 jinx_name: nql
 description: "Run NPC-SQL models with AI-powered transformations. Supports cron scheduling."
 inputs:
-  - models_dir: "~/.npcsh/npc_team/models"
-  - db: "~/npcsh_history.db"
-  - model: ""
-  - schema: ""
-  - show: ""
-  - cron: ""
-  - install_cron: ""
+- models_dir: "~/.npcsh/npc_team/models"
+- db: "~/npcsh_history.db"
+- model: ""
+- schema: ""
+- show: ""
+- cron: ""
+- install_cron: ""
 
 steps:
   - name: run_nql

npcsh/npc_team/jinxs/bin/roll.jinx

@@ -1,47 +1,44 @@
 jinx_name: "roll"
 description: "Generate a video from a text prompt."
 inputs:
-  - prompt: "" # Required text prompt for video generation.
-  - vgmodel: "" # Video generation model to use. Defaults to NPCSH_VIDEO_GEN_MODEL or NPC's model.
-  - vgprovider: "" # Video generation provider to use. Defaults to NPCSH_VIDEO_GEN_PROVIDER or NPC's provider.
-  - num_frames: 125 # Number of frames for the video.
-  - width: 256 # Width of the video.
-  - height: 256 # Height of the video.
-  - output_path: "output.mp4" # Output file path for the video.
+- prompt: ""
+- vgmodel: ""
+- vgprovider: ""
+- num_frames: 125
+- width: 256
+- height: 256
+- output_path: "output.mp4"
 steps:
   - name: "generate_video"
     engine: "python"
     code: |
       import traceback
       from npcpy.llm_funcs import gen_video
-      # Assuming NPCSH_VIDEO_GEN_MODEL and NPCSH_VIDEO_GEN_PROVIDER are accessible
-      
+
       prompt = context.get('prompt')
-      num_frames = int(context.get('num_frames', 125)) # Ensure int type
-      width = int(context.get('width', 256)) # Ensure int type
-      height = int(context.get('height', 256)) # Ensure int type
+      num_frames = int(context.get('num_frames', 125))
+      width = int(context.get('width', 256))
+      height = int(context.get('height', 256))
       output_path = context.get('output_path')
       video_gen_model = context.get('vgmodel')
       video_gen_provider = context.get('vgprovider')
       output_messages = context.get('messages', [])
       current_npc = context.get('npc')
-      
+
       if not prompt or not prompt.strip():
           context['output'] = "Usage: /roll <your prompt>"
           context['messages'] = output_messages
           exit()
 
-      # Fallback for model/provider if not explicitly set in Jinx inputs
       if not video_gen_model and current_npc and current_npc.model:
           video_gen_model = current_npc.model
       if not video_gen_provider and current_npc and current_npc.provider:
           video_gen_provider = current_npc.provider
-      
-      # Final fallbacks (these would ideally come from npcsh._state config)
-      if not video_gen_model: 
-            video_gen_model = "stable-video-diffusion" # Example default
-      if not video_gen_provider: 
-            video_gen_provider = "diffusers" # Example default
+
+      if not video_gen_model:
+            video_gen_model = "stable-video-diffusion"
+      if not video_gen_provider:
+            video_gen_provider = "diffusers"
 
       try:
           result = gen_video(
@@ -53,9 +50,9 @@ steps:
               width=width,
               height=height,
               output_path=output_path,
-              **context.get('api_kwargs', {}) # Assuming api_kwargs might be passed
+              **context.get('api_kwargs', {})
           )
-          
+
           if isinstance(result, dict):
               context['output'] = result.get('output', 'Video generated.')
               context['messages'] = result.get('messages', output_messages)
@@ -65,4 +62,4 @@ steps:
       except Exception as e:
           traceback.print_exc()
           context['output'] = f"Error generating video: {e}"
-          context['messages'] = output_messages
+          context['messages'] = output_messages

npcsh/npc_team/jinxs/bin/sample.jinx

@@ -1,9 +1,9 @@
 jinx_name: "sample"
 description: "Send a prompt directly to the LLM."
 inputs:
-  - prompt: "" # Required text prompt to send to the LLM.
-  - model: "" # LLM model to use. Defaults to NPC's model.
-  - provider: "" # LLM provider to use. Defaults to NPC's provider.
+- prompt: ""
+- model: ""
+- provider: ""
 steps:
   - name: "send_prompt_to_llm"
     engine: "python"
@@ -28,9 +28,9 @@ steps:
       if not llm_provider and current_npc and current_npc.provider:
           llm_provider = current_npc.provider
       
-      # Final fallbacks (these would ideally come from npcsh._state config)
-      if not llm_model: llm_model = "gemini-1.5-pro" # Example default
-      if not llm_provider: llm_provider = "gemini" # Example default
+      # Final fallbacks from state
+      if not llm_model: llm_model = state.chat_model if state else "llama3.2"
+      if not llm_provider: llm_provider = state.chat_provider if state else "ollama"
 
       try:
           result = get_llm_response(
@@ -38,7 +38,6 @@ steps:
               model=llm_model,
               provider=llm_provider,
               npc=current_npc,
-              **{k:v for k,v in context.items() if k not in ['messages', 'prompt', 'model', 'provider']} # Pass other context
           )
           
           if isinstance(result, dict):

npcsh/npc_team/jinxs/bin/sync.jinx

@@ -1,12 +1,12 @@
 jinx_name: "sync"
 description: "Sync npc_team files from the npcsh repo to ~/.npcsh/npc_team. Detects local modifications before overwriting."
 inputs:
-  - force: "" # Use --force or -f to overwrite all files without prompting
-  - dry_run: "" # Use --dry-run or -d to preview changes without applying them
-  - jinxs: "" # Use --jinxs to sync only .jinx files
-  - npcs: "" # Use --npcs to sync only .npc files
-  - ctx: "" # Use --ctx to sync only .ctx files
-  - images: "" # Use --images to sync only image files (.png, .jpg, .jpeg)
+- force: ""
+- dry_run: ""
+- jinxs: ""
+- npcs: ""
+- ctx: ""
+- images: ""
 steps:
   - name: "sync_npc_team"
     engine: "python"

npcsh/npc_team/jinxs/bin/vixynt.jinx

@@ -1,14 +1,14 @@
 jinx_name: "vixynt"
 description: "Generates images from text descriptions or edits existing ones."
 inputs:
-  - prompt
-  - model: null
-  - provider: null
-  - output_name: null
-  - attachments: null
-  - n_images: null
-  - height: null
-  - width: null
+- prompt
+- model: null
+- provider: null
+- output_name: null
+- attachments: null
+- n_images: null
+- height: null
+- width: null
 steps:
   - name: "generate_or_edit_image"
     engine: "python"

npcsh/npc_team/jinxs/incognide/add_tab.jinx

@@ -0,0 +1,11 @@
+jinx_name: studio.add_tab
+description: Add a new tab to a pane.
+inputs:
+- paneId: "active"
+- type: ""
+- path: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/close_pane.jinx

@@ -0,0 +1,9 @@
+jinx_name: studio.close_pane
+description: Close a pane in NPC Studio. Use paneId="active" or omit to close the active pane.
+inputs:
+- paneId: "active"
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/close_tab.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.close_tab
+description: Close a specific tab in a pane.
+inputs:
+- paneId: "active"
+- tabIndex: 0
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/confirm.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.confirm
+description: Show a confirmation dialog and return the user's choice.
+inputs:
+- message: ""
+- title: "Confirm"
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/focus_pane.jinx

@@ -0,0 +1,9 @@
+jinx_name: studio.focus_pane
+description: Focus/activate a specific pane in NPC Studio.
+inputs:
+- paneId: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/{npc_studio/npc-studio.jinx → incognide/incognide.jinx}

@@ -1,7 +1,7 @@
-jinx_name: "npc-studio"
+jinx_name: "incognide"
 description: "Start npc studio"
 inputs:
-  - user_command: ""
+- user_command: ""
 steps:
   - name: "launch_npc_studio"
     engine: "python"

npcsh/npc_team/jinxs/incognide/list_panes.jinx

@@ -0,0 +1,8 @@
+jinx_name: studio.list_panes
+description: List all open panes in NPC Studio. Returns pane IDs, types, titles, and which is active.
+inputs: []
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/navigate.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.navigate
+description: Navigate a browser pane to a specific URL.
+inputs:
+- paneId: "active"
+- url: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/notify.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.notify
+description: Show a notification toast in NPC Studio.
+inputs:
+- message: ""
+- type: "info"
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/open_pane.jinx

@@ -0,0 +1,13 @@
+jinx_name: studio.open_pane
+description: Open a new pane in NPC Studio. Supports editor, terminal, browser, pdf, csv, chat, image, folder, and other content types.
+inputs:
+- type: ""
+- path: ""
+- position: "right"
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      # This action is executed by the NPC Studio frontend
+      # The frontend intercepts studio.* tool calls and handles them directly
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/read_pane.jinx

@@ -0,0 +1,9 @@
+jinx_name: studio.read_pane
+description: Read the contents of a pane. For editor panes returns file content, for chat panes returns messages, for browser panes returns URL/title.
+inputs:
+- paneId: "active"
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/run_terminal.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.run_terminal
+description: Execute a command in a terminal pane.
+inputs:
+- paneId: "active"
+- command: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/send_message.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.send_message
+description: Send a message in a chat pane.
+inputs:
+- paneId: "active"
+- message: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/split_pane.jinx

@@ -0,0 +1,12 @@
+jinx_name: studio.split_pane
+description: Split an existing pane to create a new pane alongside it.
+inputs:
+- paneId: "active"
+- direction: "right"
+- type: ""
+- path: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/switch_npc.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.switch_npc
+description: Switch the active NPC in a chat pane.
+inputs:
+- paneId: "active"
+- npcName: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/switch_tab.jinx

@@ -0,0 +1,10 @@
+jinx_name: studio.switch_tab
+description: Switch to a specific tab in a pane.
+inputs:
+- paneId: "active"
+- tabIndex: 0
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/write_file.jinx

@@ -0,0 +1,11 @@
+jinx_name: studio.write_file
+description: Write content to an editor pane. Updates the file content in the pane.
+inputs:
+- paneId: "active"
+- content: ""
+- path: ""
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/incognide/zen_mode.jinx

@@ -0,0 +1,9 @@
+jinx_name: studio.zen_mode
+description: Toggle zen mode (fullscreen) for a pane.
+inputs:
+- paneId: "active"
+steps:
+  - name: frontend_action
+    engine: python
+    code: |
+      context['output'] = "Action executed by frontend"

npcsh/npc_team/jinxs/lib/browser/browser_action.jinx

@@ -12,14 +12,14 @@ description: |
   - get_page: Get page title, URL, and visible text
   - get_elements: Get interactive elements with their selectors
   - press_key: Press a key (enter, tab, escape, etc)
-  Selectors: CSS (#id, .class, input[name="x"]) or xpath://... for XPath
+  Selectors: CSS (
 inputs:
-  - action:
+- action:
       description: "Action: click, type, type_and_enter, set_value, select, wait, scroll, get_text, get_page, get_elements, press_key"
-  - selector:
+- selector:
       description: "CSS selector or XPath (prefix xpath: for XPath)"
       default: ""
-  - value:
+- value:
       description: "Value for type/select, or scroll direction, or key name"
       default: ""
 

npcsh/npc_team/jinxs/lib/browser/browser_screenshot.jinx

@@ -1,7 +1,7 @@
 jinx_name: browser_screenshot
 description: Take a screenshot of the current browser page.
 inputs:
-  - filename:
+- filename:
       description: "Optional filename for screenshot"
       default: ""
 

npcsh/npc_team/jinxs/lib/browser/open_browser.jinx

@@ -3,9 +3,9 @@ description: |
   Open a browser and navigate to a URL. The browser stays open for follow-up commands.
   Use this to start browser automation.
 inputs:
-  - url:
+- url:
       description: "URL to navigate to"
-  - browser: "firefox"
+- browser: "firefox"
 
 steps:
   - name: open_browser

npcsh/npc_team/jinxs/lib/computer_use/click.jinx

@@ -1,8 +1,8 @@
 jinx_name: click
 description: Click at screen coordinates (0-100 percentage)
 inputs:
-  - x: 50  # X coordinate as percentage (0-100)
-  - y: 50  # Y coordinate as percentage (0-100)
+- x: 50
+- y: 50
 
 steps:
   - name: perform_click

npcsh/npc_team/jinxs/lib/computer_use/key_press.jinx

@@ -6,7 +6,7 @@ description: |
   For combinations use + like: ctrl+a, ctrl+c, ctrl+v, alt+tab, ctrl+shift+t
   For regular letters/numbers, use type_text instead.
 inputs:
-  - key: "enter"
+- key: "enter"
 
 steps:
   - name: perform_key

npcsh/npc_team/jinxs/lib/computer_use/launch_app.jinx

@@ -1,7 +1,7 @@
 jinx_name: launch_app
 description: Launch an application on the system
 inputs:
-  - command: ""  # Command to launch (e.g., "open -a Firefox" on macOS)
+- command: ""
 
 steps:
   - name: perform_launch

npcsh/npc_team/jinxs/lib/computer_use/screenshot.jinx

@@ -1,7 +1,7 @@
 jinx_name: screenshot
 description: Capture a screenshot of the current screen
 inputs:
-  - output_path: null  # Optional path to save screenshot
+- output_path: null
 
 steps:
   - name: capture_screenshot

npcsh/npc_team/jinxs/lib/computer_use/trigger.jinx

@@ -1,8 +1,8 @@
 jinx_name: "trigger"
 description: "Creates a persistent listener (--listen) or a scheduled task (--cron)."
 inputs:
-  - listen: "" # The description for a persistent, event-driven listener.
-  - cron: ""   # The description for a scheduled, time-based task.
+- listen: ""
+- cron: ""
 steps:
   - name: "execute_command"
     engine: "python"

npcsh/npc_team/jinxs/lib/computer_use/type_text.jinx

@@ -1,7 +1,7 @@
 jinx_name: type_text
 description: Type text using keyboard
 inputs:
-  - text: ""  # Text to type
+- text: ""
 
 steps:
   - name: perform_type

npcsh/npc_team/jinxs/lib/computer_use/wait.jinx

@@ -1,7 +1,7 @@
 jinx_name: wait
 description: Wait/pause for a specified duration in seconds
 inputs:
-  - duration: 1  # Duration to wait in seconds
+- duration: 1
 
 steps:
   - name: perform_wait