npcsh 1.1.13__py3-none-any.whl → 1.1.15__py3-none-any.whl

npcsh/npc_team/jinxs/modes/pti.jinx

@@ -1,170 +0,0 @@
-jinx_name: pti
-description: Pardon-The-Interruption - human-in-the-loop reasoning with think tags and interruptible streaming
-npc: frederic
-inputs:
-  - model: null
-  - provider: null
-  - files: null
-  - reasoning_model: null
-
-steps:
-  - name: pti_repl
-    engine: python
-    code: |
-      import os
-      import sys
-      from termcolor import colored
-
-      from npcpy.llm_funcs import get_llm_response
-      from npcpy.npc_sysenv import get_system_message, render_markdown
-      from npcpy.data.load import load_file_contents
-      from npcpy.data.text import rag_search
-
-      npc = context.get('npc')
-      team = context.get('team')
-      messages = context.get('messages', [])
-      files = context.get('files')
-
-      # PTI uses reasoning model for deeper thinking
-      model = context.get('reasoning_model') or context.get('model') or (npc.model if npc else None)
-      provider = context.get('provider') or (npc.provider if npc else None)
-
-      print("""
-      ██████╗ ████████╗██╗
-      ██╔══██╗╚══██╔══╝██║
-      ██████╔╝   ██║   ██║
-      ██╔═══╝    ██║   ██║
-      ██║        ██║   ██║
-      ╚═╝        ╚═╝   ╚═╝
-
-      Pardon-The-Interruption
-      Human-in-the-loop reasoning mode
-      """)
-
-      npc_name = npc.name if npc else "pti"
-      print(f"Entering PTI mode (NPC: {npc_name}). Type '/pq' to exit.")
-      print("  - AI will use <think> tags for step-by-step reasoning")
-      print("  - Use <request_for_input> to pause and ask questions")
-      print("  - Ctrl+C interrupts stream for immediate feedback")
-
-      # Load files if provided
-      loaded_content = {}
-      if files:
-          if isinstance(files, str):
-              files = [f.strip() for f in files.split(',')]
-          for file_path in files:
-              file_path = os.path.expanduser(file_path)
-              if os.path.exists(file_path):
-                  try:
-                      chunks = load_file_contents(file_path)
-                      loaded_content[file_path] = "\n".join(chunks)
-                      print(colored(f"Loaded: {file_path}", "green"))
-                  except Exception as e:
-                      print(colored(f"Error loading {file_path}: {e}", "red"))
-
-      # System message for PTI mode
-      pti_system = """You are an AI assistant in PTI (Pardon-The-Interruption) mode.
-
-      IMPORTANT INSTRUCTIONS:
-      1. Think step-by-step using <think>...</think> tags to show your reasoning
-      2. When you need more information from the user, use <request_for_input>your question</request_for_input>
-      3. Be thorough but concise in your reasoning
-      4. The user can interrupt at any time to provide guidance
-
-      Example:
-      <think>
-      Let me break this down...
-      Step 1: First I need to understand X
-      Step 2: Then consider Y
-      </think>
-
-      <request_for_input>
-      I notice you mentioned Z. Could you clarify what you mean by that?
-      </request_for_input>"""
-
-      if not messages or messages[0].get("role") != "system":
-          messages.insert(0, {"role": "system", "content": pti_system})
-
-      # REPL loop
-      user_input = None
-      while True:
-          try:
-              if not user_input:
-                  prompt_str = f"{npc_name}:pti> "
-                  user_input = input(prompt_str).strip()
-
-              if not user_input:
-                  user_input = None
-                  continue
-
-              if user_input.lower() == "/pq":
-                  print("Exiting PTI mode.")
-                  break
-
-              # Build prompt with file context
-              prompt_for_llm = user_input
-              if loaded_content:
-                  context_str = "\n".join([f"--- {fname} ---\n{content}" for fname, content in loaded_content.items()])
-                  prompt_for_llm += f"\n\nContext:\n{context_str}"
-
-              prompt_for_llm += "\n\nThink step-by-step using <think> tags. Use <request_for_input> when you need clarification."
-
-              messages.append({"role": "user", "content": user_input})
-
-              try:
-                  resp = get_llm_response(
-                      prompt_for_llm,
-                      model=model,
-                      provider=provider,
-                      messages=messages[:-1],  # Don't duplicate the user message
-                      stream=True,
-                      npc=npc
-                  )
-
-                  response_stream = resp.get('response')
-                  full_response = ""
-                  request_found = False
-
-                  # Stream the response
-                  for chunk in response_stream:
-                      chunk_content = ""
-                      if hasattr(chunk, 'choices') and chunk.choices:
-                          delta = chunk.choices[0].delta
-                          if hasattr(delta, 'content') and delta.content:
-                              chunk_content = delta.content
-                      elif isinstance(chunk, dict):
-                          chunk_content = chunk.get("message", {}).get("content", "")
-
-                      if chunk_content:
-                          print(chunk_content, end='', flush=True)
-                          full_response += chunk_content
-
-                          # Check for request_for_input
-                          if "</request_for_input>" in full_response:
-                              request_found = True
-                              break
-
-                  print()  # newline after stream
-
-                  messages.append({"role": "assistant", "content": full_response})
-                  user_input = None  # Reset for next iteration
-
-              except KeyboardInterrupt:
-                  print(colored("\n\n--- Interrupted ---", "yellow"))
-                  interrupt_input = input("Your feedback: ").strip()
-                  if interrupt_input:
-                      user_input = interrupt_input
-                  else:
-                      user_input = None
-                  continue
-
-          except KeyboardInterrupt:
-              print("\nUse '/pq' to exit or continue.")
-              user_input = None
-              continue
-          except EOFError:
-              print("\nExiting PTI mode.")
-              break
-
-      context['output'] = "Exited PTI mode."
-      context['messages'] = messages

npcsh/npc_team/jinxs/modes/wander.jinx

@@ -1,186 +0,0 @@
-jinx_name: wander
-description: Experimental wandering mode - creative exploration with varied temperatures and random events
-inputs:
-  - problem: null
-  - environment: null
-  - low_temp: 0.5
-  - high_temp: 1.9
-  - interruption_likelihood: 1.0
-  - sample_rate: 0.4
-  - n_streams: 5
-  - include_events: false
-  - num_events: 3
-  - model: null
-  - provider: null
-
-steps:
-  - name: wander_explore
-    engine: python
-    code: |
-      import os
-      import random
-      from termcolor import colored
-
-      from npcpy.llm_funcs import get_llm_response
-
-      npc = context.get('npc')
-      messages = context.get('messages', [])
-
-      problem = context.get('problem')
-      environment = context.get('environment')
-      low_temp = float(context.get('low_temp', 0.5))
-      high_temp = float(context.get('high_temp', 1.9))
-      interruption_likelihood = float(context.get('interruption_likelihood', 1.0))
-      sample_rate = float(context.get('sample_rate', 0.4))
-      n_streams = int(context.get('n_streams', 5))
-      include_events = context.get('include_events', False)
-      num_events = int(context.get('num_events', 3))
-
-      model = context.get('model') or (npc.model if npc else 'gemini-1.5-pro')
-      provider = context.get('provider') or (npc.provider if npc else 'gemini')
-
-      if not problem:
-          context['output'] = """Usage: /wander <problem to explore>
-
-      Options:
-        --environment DESC   Metaphorical environment for wandering
-        --low-temp F        Low temperature (default: 0.5)
-        --high-temp F       High temperature (default: 1.9)
-        --n-streams N       Number of exploration streams (default: 5)
-        --include-events    Add random events during wandering
-
-      Example: /wander How might we reimagine urban transportation?"""
-          context['messages'] = messages
-          exit()
-
-      print(f"""
-      ██╗    ██╗ █████╗ ███╗   ██╗██████╗ ███████╗██████╗
-      ██║    ██║██╔══██╗████╗  ██║██╔══██╗██╔════╝██╔══██╗
-      ██║ █╗ ██║███████║██╔██╗ ██║██║  ██║█████╗  ██████╔╝
-      ██║███╗██║██╔══██║██║╚██╗██║██║  ██║██╔══╝  ██╔══██╗
-      ╚███╔███╔╝██║  ██║██║ ╚████║██████╔╝███████╗██║  ██║
-       ╚══╝╚══╝ ╚═╝  ╚═╝╚═╝  ╚═══╝╚═════╝ ╚══════╝╚═╝  ╚═╝
-
-      Experimental Wandering Mode
-      Problem: {problem}
-      Temperature range: {low_temp} - {high_temp}
-      Streams: {n_streams}
-      """)
-
-      # Generate environment if not provided
-      if not environment:
-          env_prompt = f"""Create a rich, metaphorical environment for wandering through while thinking about:
-          "{problem}"
-
-          The environment should:
-          1. Have distinct regions or areas
-          2. Include various elements and features
-          3. Be metaphorically related to the problem
-          4. Be described in 3-5 sentences
-
-          Provide only the description, no framing."""
-
-          print(colored("Generating wandering environment...", "cyan"))
-          resp = get_llm_response(env_prompt, model=model, provider=provider, temperature=0.7, npc=npc)
-          environment = str(resp.get('response', 'A vast conceptual landscape stretches before you.'))
-          print(f"\n{environment}\n")
-
-      # Event types for random encounters
-      event_types = ["encounter", "discovery", "obstacle", "insight", "shift", "memory"]
-
-      all_insights = []
-      wandering_log = []
-
-      for stream_idx in range(n_streams):
-          # Alternate between low and high temperature
-          if stream_idx % 2 == 0:
-              temp = low_temp
-              mode = "focused"
-          else:
-              temp = high_temp
-              mode = "creative"
-
-          print(colored(f"\n--- Stream {stream_idx + 1}/{n_streams} ({mode}, temp={temp}) ---", "cyan"))
-
-          # Generate random event if enabled
-          event_context = ""
-          if include_events and random.random() < sample_rate:
-              event_type = random.choice(event_types)
-              event_prompt = f"""In the environment: {environment}
-
-              While exploring the problem "{problem}", generate a {event_type} event.
-              The event should be metaphorical and relate to the problem.
-              Describe it in 2-3 sentences."""
-
-              event_resp = get_llm_response(event_prompt, model=model, provider=provider, temperature=0.9, npc=npc)
-              event = str(event_resp.get('response', ''))
-              event_context = f"\n\nEvent ({event_type}): {event}"
-              print(colored(f"[{event_type.upper()}] {event[:100]}...", "yellow"))
-
-          # Main wandering exploration
-          wander_prompt = f"""You are wandering through: {environment}
-
-          Problem being explored: "{problem}"
-          {event_context}
-
-          Previous insights: {all_insights[-3:] if all_insights else 'Starting fresh'}
-
-          In this {mode} exploration (temperature {temp}):
-          1. Let your mind wander through the conceptual space
-          2. Make unexpected connections
-          3. Notice what emerges from the wandering
-          4. Share any insights, questions, or realizations
-
-          Think freely and explore."""
-
-          resp = get_llm_response(wander_prompt, model=model, provider=provider, temperature=temp, npc=npc)
-          stream_output = str(resp.get('response', ''))
-          print(stream_output)
-
-          all_insights.append(stream_output)
-          wandering_log.append({
-              "stream": stream_idx + 1,
-              "mode": mode,
-              "temperature": temp,
-              "event": event_context if include_events else None,
-              "insight": stream_output
-          })
-
-          # Random interruption
-          if random.random() < interruption_likelihood * 0.2:
-              print(colored("\n[Pause for reflection...]", "magenta"))
-              reflect_prompt = f"Briefly reflect on what's emerged so far about: {problem}"
-              reflect_resp = get_llm_response(reflect_prompt, model=model, provider=provider, temperature=0.4, npc=npc)
-              print(colored(str(reflect_resp.get('response', ''))[:200], "magenta"))
-
-      # Synthesis
-      print(colored("\n--- Synthesizing Wanderings ---", "cyan"))
-
-      synthesis_prompt = f"""After wandering through "{environment}" exploring "{problem}":
-
-      All insights gathered:
-      {chr(10).join(all_insights)}
-
-      Synthesize what emerged from this wandering:
-      1. Key themes that appeared
-      2. Unexpected connections made
-      3. New questions raised
-      4. Potential directions to explore further"""
-
-      resp = get_llm_response(synthesis_prompt, model=model, provider=provider, temperature=0.5, npc=npc)
-      synthesis = str(resp.get('response', ''))
-
-      print("\n" + "="*50)
-      print(colored("WANDERING SYNTHESIS", "green", attrs=['bold']))
-      print("="*50)
-      print(synthesis)
-
-      context['output'] = synthesis
-      context['messages'] = messages
-      context['wander_result'] = {
-          'problem': problem,
-          'environment': environment,
-          'log': wandering_log,
-          'insights': all_insights,
-          'synthesis': synthesis
-      }

npcsh/npc_team/jinxs/utils/agent.jinx

@@ -1,17 +0,0 @@
-jinx_name: agent
-description: Provides an LLM response with tool use enabled.
-inputs:
-  - query
-  - auto_process_tool_calls: True
-  - use_core_tools: True
-steps:
-  - name: get_agent_response
-    engine: python
-    code: |
-      response = npc.get_llm_response(
-          request=query,
-          messages=context.get('messages', []),
-          auto_process_tool_calls={{ auto_process_tool_calls | default(True) }},
-          use_core_tools={{ use_core_tools | default(True) }}
-      )
-      output = response.get('response', '')

npcsh/npc_team/jinxs/utils/core/jinxs.jinx

@@ -1,32 +0,0 @@
-jinx_name: jinxs
-description: Show available jinxs for the current NPC/Team
-inputs: []
-steps:
-  - name: list_jinxs
-    engine: python
-    code: |
-      output_lines = ["Available Jinxs:\n"]
-      jinxs_listed = set()
-      
-      if hasattr(npc, 'team') and npc.team:
-          team = npc.team
-          
-          if hasattr(team, 'jinxs_dict') and team.jinxs_dict:
-              output_lines.append(f"\n--- Team Jinxs ---\n")
-              for name, jinx_obj in sorted(team.jinxs_dict.items()):
-                  desc = getattr(jinx_obj, 'description', 'No description available.')
-                  output_lines.append(f"- /{name}: {desc}\n")
-                  jinxs_listed.add(name)
-      
-      if hasattr(npc, 'jinxs_dict') and npc.jinxs_dict:
-          output_lines.append(f"\n--- NPC Jinxs for {npc.name} ---\n")
-          for name, jinx_obj in sorted(npc.jinxs_dict.items()):
-              if name not in jinxs_listed:
-                  desc = getattr(jinx_obj, 'description', 'No description available.')
-                  output_lines.append(f"- /{name}: {desc}\n")
-                  jinxs_listed.add(name)
-      
-      if not jinxs_listed:
-          output = "No jinxs available for the current context."
-      else:
-          output = "".join(output_lines)

npcsh-1.1.13.data/data/npcsh/npc_team/agent.jinx

@@ -1,17 +0,0 @@
-jinx_name: agent
-description: Provides an LLM response with tool use enabled.
-inputs:
-  - query
-  - auto_process_tool_calls: True
-  - use_core_tools: True
-steps:
-  - name: get_agent_response
-    engine: python
-    code: |
-      response = npc.get_llm_response(
-          request=query,
-          messages=context.get('messages', []),
-          auto_process_tool_calls={{ auto_process_tool_calls | default(True) }},
-          use_core_tools={{ use_core_tools | default(True) }}
-      )
-      output = response.get('response', '')

npcsh-1.1.13.data/data/npcsh/npc_team/alicanto.jinx

@@ -1,194 +0,0 @@
-jinx_name: alicanto
-description: Deep research mode - multi-perspective exploration with gold insights and cliff warnings
-npc: forenpc
-inputs:
-  - query: null
-  - num_npcs: 5
-  - depth: 3
-  - model: null
-  - provider: null
-  - max_steps: 20
-  - skip_research: true
-  - exploration: 0.3
-  - creativity: 0.5
-  - format: report
-
-steps:
-  - name: alicanto_research
-    engine: python
-    code: |
-      import os
-      from termcolor import colored
-
-      from npcpy.llm_funcs import get_llm_response
-      from npcpy.data.web import search_web
-      from npcpy.npc_compiler import NPC
-
-      npc = context.get('npc')
-      team = context.get('team')
-      messages = context.get('messages', [])
-
-      query = context.get('query')
-      num_npcs = int(context.get('num_npcs', 5))
-      depth = int(context.get('depth', 3))
-      max_steps = int(context.get('max_steps', 20))
-      skip_research = context.get('skip_research', True)
-      exploration = float(context.get('exploration', 0.3))
-      creativity = float(context.get('creativity', 0.5))
-      output_format = context.get('format', 'report')
-
-      model = context.get('model') or (npc.model if npc else 'gemini-1.5-pro')
-      provider = context.get('provider') or (npc.provider if npc else 'gemini')
-
-      if not query:
-          context['output'] = """Usage: /alicanto <research query>
-
-      Options:
-        --num-npcs N      Number of research perspectives (default: 5)
-        --depth N         Research depth (default: 3)
-        --max-steps N     Maximum research steps (default: 20)
-        --exploration F   Exploration factor 0-1 (default: 0.3)
-        --creativity F    Creativity factor 0-1 (default: 0.5)
-        --format FORMAT   Output: report|summary|full (default: report)
-
-      Example: /alicanto What are the latest advances in quantum computing?"""
-          context['messages'] = messages
-          exit()
-
-      print(f"""
-       █████╗ ██╗     ██╗ ██████╗ █████╗ ███╗   ██╗████████╗ ██████╗
-      ██╔══██╗██║     ██║██╔════╝██╔══██╗████╗  ██║╚══██╔══╝██╔═══██╗
-      ███████║██║     ██║██║     ███████║██╔██╗ ██║   ██║   ██║   ██║
-      ██╔══██║██║     ██║██║     ██╔══██║██║╚██╗██║   ██║   ██║   ██║
-      ██║  ██║███████╗██║╚██████╗██║  ██║██║ ╚████║   ██║   ╚██████╔╝
-      ╚═╝  ╚═╝╚══════╝╚═╝ ╚═════╝╚═╝  ╚═╝╚═╝  ╚═══╝   ╚═╝    ╚═════╝
-
-      Deep Research Mode
-      Query: {query}
-      Perspectives: {num_npcs} | Depth: {depth} | Max Steps: {max_steps}
-      """)
-
-      # Generate research perspectives
-      perspectives_prompt = f"""Generate {num_npcs} distinct research perspectives for investigating: "{query}"
-
-      For each perspective, provide:
-      1. Name (a descriptive title)
-      2. Approach (how this perspective would investigate)
-      3. Key questions to explore
-
-      Return as a numbered list."""
-
-      print(colored("Generating research perspectives...", "cyan"))
-      resp = get_llm_response(
-          perspectives_prompt,
-          model=model,
-          provider=provider,
-          npc=npc
-      )
-      perspectives = str(resp.get('response', ''))
-      print(perspectives)
-
-      # Conduct web research if not skipped
-      research_findings = ""
-      if not skip_research:
-          print(colored("\nConducting web research...", "cyan"))
-          try:
-              search_results = search_web(query, n_results=5)
-              if search_results:
-                  research_findings = "\n\nWeb Research Findings:\n"
-                  for i, result in enumerate(search_results[:5], 1):
-                      title = result.get('title', 'No title')
-                      snippet = result.get('snippet', result.get('body', ''))[:200]
-                      research_findings += f"\n{i}. {title}\n   {snippet}...\n"
-                  print(colored(f"Found {len(search_results)} sources", "green"))
-          except Exception as e:
-              print(colored(f"Web search error: {e}", "yellow"))
-
-      # Multi-step exploration from each perspective
-      all_insights = []
-      gold_insights = []  # Key valuable findings
-      cliff_warnings = []  # Potential pitfalls or caveats
-
-      for step in range(min(depth, max_steps)):
-          print(colored(f"\n--- Research Depth {step + 1}/{depth} ---", "cyan"))
-
-          explore_prompt = f"""Research query: "{query}"
-
-      Perspectives generated:
-      {perspectives}
-
-      {research_findings}
-
-      Previous insights: {all_insights[-3:] if all_insights else 'None yet'}
-
-      For depth level {step + 1}:
-      1. Explore deeper implications from each perspective
-      2. Identify GOLD insights (valuable, non-obvious findings) - mark with [GOLD]
-      3. Identify CLIFF warnings (pitfalls, caveats, risks) - mark with [CLIFF]
-      4. Connect insights across perspectives
-
-      Exploration factor: {exploration} (higher = more diverse exploration)
-      Creativity factor: {creativity} (higher = more novel connections)"""
-
-          resp = get_llm_response(
-              explore_prompt,
-              model=model,
-              provider=provider,
-              temperature=creativity,
-              npc=npc
-          )
-
-          step_insights = str(resp.get('response', ''))
-          print(step_insights)
-
-          # Extract gold and cliff markers
-          if '[GOLD]' in step_insights:
-              gold_insights.extend([line.strip() for line in step_insights.split('\n') if '[GOLD]' in line])
-          if '[CLIFF]' in step_insights:
-              cliff_warnings.extend([line.strip() for line in step_insights.split('\n') if '[CLIFF]' in line])
-
-          all_insights.append(step_insights)
-
-      # Generate final synthesis
-      print(colored("\n--- Synthesizing Research ---", "cyan"))
-
-      synthesis_prompt = f"""Synthesize research on: "{query}"
-
-      All insights gathered:
-      {chr(10).join(all_insights)}
-
-      Gold insights identified:
-      {chr(10).join(gold_insights) if gold_insights else 'None explicitly marked'}
-
-      Cliff warnings identified:
-      {chr(10).join(cliff_warnings) if cliff_warnings else 'None explicitly marked'}
-
-      Generate a {output_format} that:
-      1. Summarizes key findings
-      2. Highlights the most valuable insights (gold)
-      3. Notes important caveats and risks (cliffs)
-      4. Provides actionable conclusions"""
-
-      resp = get_llm_response(
-          synthesis_prompt,
-          model=model,
-          provider=provider,
-          npc=npc
-      )
-
-      final_report = str(resp.get('response', ''))
-      print("\n" + "="*60)
-      print(colored("ALICANTO RESEARCH REPORT", "green", attrs=['bold']))
-      print("="*60)
-      print(final_report)
-
-      context['output'] = final_report
-      context['messages'] = messages
-      context['alicanto_result'] = {
-          'query': query,
-          'perspectives': perspectives,
-          'insights': all_insights,
-          'gold': gold_insights,
-          'cliffs': cliff_warnings,
-          'report': final_report
-      }

npcsh-1.1.13.data/data/npcsh/npc_team/alicanto.npc

@@ -1,2 +0,0 @@
-name: alicanto
-primary_directive: You are Alicanto the mythical bird. You have been spotted and it is your job to lead users to explore the world.