npcsh 0.3.29__tar.gz → 0.3.31__tar.gz

{npcsh-0.3.29/npcsh.egg-info → npcsh-0.3.31}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: npcsh
-Version: 0.3.29
+Version: 0.3.31
 Summary: npcsh is a command line tool for integrating LLMs into everyday workflows and for orchestrating teams of NPCs.
 Home-page: https://github.com/cagostino/npcsh
 Author: Christopher Agostino
@@ -112,22 +112,23 @@ Interested to stay in the loop and to hear the latest and greatest about `npcsh`
 ## TLDR Cheat Sheet
 Users can take advantage of `npcsh` through its custom shell or through a command-line interface (CLI) tool. Below is a cheat sheet that shows how to use `npcsh` commands in both the shell and the CLI. For the npcsh commands to work, one must activate `npcsh` by typing it in a shell.
 
+
+
+
 | Task | npc CLI | npcsh |
 |----------|----------|----------|
 | Ask a generic question | npc 'prompt' | 'prompt' |
 | Compile an NPC | npc compile /path/to/npc.npc | /compile /path/to/npc.npc |
 | Computer use | npc plonk -n 'npc_name' -sp 'task for plonk to carry out '| /plonk -n 'npc_name' -sp 'task for plonk to carry out ' |
 | Conjure an NPC team from context and templates | npc init -t 'template1, template2' -ctx 'context'   | /conjure  -t 'template1, 'template2' -ctx 'context'  |
-| Enter a chat with an NPC (NPC needs to be compiled first) | npc npc_name | /npc_name |
+| Enter a chat with an NPC (NPC needs to be compiled first) | npc chat -n npc_name | /spool npc=<npc_name> |
 | Generate image    | npc vixynt 'prompt'  | /vixynt prompt   |
 | Get a sample LLM response  | npc sample 'prompt'   | /sample prompt for llm  |
-| Invoke a tool  | npc tool {tool_name} -args --flags | /tool_name -args --flags |
-| Search locally | npc tool local_search -args --flags | /local_search -args --flags |
 | Search for a term in the npcsh_db only in conversations with a specific npc | npc rag -n 'npc_name' -f 'filename' -q 'query' | /rag -n 'npc_name' -f 'filename' -q 'query' |
-| Search the web | npc search -p provider 'query' | /search -p provider 'query' |
+| Search the web | npc search -q "cal golden bears football schedule" -sp perplexity | /search -p perplexity 'cal bears football schedule' |
 | Serve an NPC team | npc serve --port 5337 --cors='http://localhost:5137/' | /serve --port 5337 --cors='http://localhost:5137/' |
 | Screenshot analysis  | npc ots |  /ots  |
-| Voice Chat    | npc whisper 'npc_name'   | /whisper   |
+| Voice Chat    | npc whisper -n 'npc_name'   | /whisper   |
 
 
 ## Python Examples
@@ -725,19 +726,6 @@ The code in the visible section of your VS Code window appears to be a script fo
 In summary, this code automates the process of capturing a screenshot, saving it with a unique filename, and analyzing that image for specific content or patterns.
 
 
-```
-
-```npcsh
-npcsh> What is the biggest file in my current folder?
-
-LLM suggests the following bash command: ls -S | head -n 1
-
-Running command: ls -S | head -n 1
-
-Command executed with output: image_20241111_000033.png
-
-I ran the command ls -S | head -n 1 in your current folder. This command sorts all files by size in descending order and then selects the first entry, which represents the largest file. The result of this operation shows that the biggest file in your current folder is image_20241111_000033.png.
-
 ```
 
 ```npcsh
@@ -869,7 +857,7 @@ and then the associated image :
 An important facet that makes `npcsh` so powerful is the ability to pipe outputs from one tool call to another. This allows for the chaining of commands and the creation of complex workflows. For example, you can use the output of a search to generate an image, or you can use the output of an image analysis to generate a report. Here is an example of how this might look in practice:
 ```npcsh
 npcsh> what is the gdp of russia in 2024? | /vixynt 'generate an image that contains {0}'
-
+```
 ### Executing Bash Commands
 You can execute bash commands directly within npcsh. The LLM can also generate and execute bash commands based on your natural language requests.
 For example:
@@ -1167,7 +1155,10 @@ Search can be accomplished through the `/search` macro. You can specify the prov
 you must set a perplexity api key as an environment variable as described above. The default provider is duckduckgo.
 
 NOTE: while google is an available search engine, they recently implemented changes (early 2025) that make the python google search package no longer as reliable.
-For now, we will use duckduckgo and revisit this issue when other more critical aspects are handled.
+Duckduckgo's search toool also givies rate limit errors often, so until a more robust
+solution is implemented for it, Perplexity's will be the most reliable.
+
+
 
 
 ```npcsh

{npcsh-0.3.29 → npcsh-0.3.31}/README.md

@@ -28,22 +28,23 @@ Interested to stay in the loop and to hear the latest and greatest about `npcsh`
 ## TLDR Cheat Sheet
 Users can take advantage of `npcsh` through its custom shell or through a command-line interface (CLI) tool. Below is a cheat sheet that shows how to use `npcsh` commands in both the shell and the CLI. For the npcsh commands to work, one must activate `npcsh` by typing it in a shell.
 
+
+
+
 | Task | npc CLI | npcsh |
 |----------|----------|----------|
 | Ask a generic question | npc 'prompt' | 'prompt' |
 | Compile an NPC | npc compile /path/to/npc.npc | /compile /path/to/npc.npc |
 | Computer use | npc plonk -n 'npc_name' -sp 'task for plonk to carry out '| /plonk -n 'npc_name' -sp 'task for plonk to carry out ' |
 | Conjure an NPC team from context and templates | npc init -t 'template1, template2' -ctx 'context'   | /conjure  -t 'template1, 'template2' -ctx 'context'  |
-| Enter a chat with an NPC (NPC needs to be compiled first) | npc npc_name | /npc_name |
+| Enter a chat with an NPC (NPC needs to be compiled first) | npc chat -n npc_name | /spool npc=<npc_name> |
 | Generate image    | npc vixynt 'prompt'  | /vixynt prompt   |
 | Get a sample LLM response  | npc sample 'prompt'   | /sample prompt for llm  |
-| Invoke a tool  | npc tool {tool_name} -args --flags | /tool_name -args --flags |
-| Search locally | npc tool local_search -args --flags | /local_search -args --flags |
 | Search for a term in the npcsh_db only in conversations with a specific npc | npc rag -n 'npc_name' -f 'filename' -q 'query' | /rag -n 'npc_name' -f 'filename' -q 'query' |
-| Search the web | npc search -p provider 'query' | /search -p provider 'query' |
+| Search the web | npc search -q "cal golden bears football schedule" -sp perplexity | /search -p perplexity 'cal bears football schedule' |
 | Serve an NPC team | npc serve --port 5337 --cors='http://localhost:5137/' | /serve --port 5337 --cors='http://localhost:5137/' |
 | Screenshot analysis  | npc ots |  /ots  |
-| Voice Chat    | npc whisper 'npc_name'   | /whisper   |
+| Voice Chat    | npc whisper -n 'npc_name'   | /whisper   |
 
 
 ## Python Examples
@@ -641,19 +642,6 @@ The code in the visible section of your VS Code window appears to be a script fo
 In summary, this code automates the process of capturing a screenshot, saving it with a unique filename, and analyzing that image for specific content or patterns.
 
 
-```
-
-```npcsh
-npcsh> What is the biggest file in my current folder?
-
-LLM suggests the following bash command: ls -S | head -n 1
-
-Running command: ls -S | head -n 1
-
-Command executed with output: image_20241111_000033.png
-
-I ran the command ls -S | head -n 1 in your current folder. This command sorts all files by size in descending order and then selects the first entry, which represents the largest file. The result of this operation shows that the biggest file in your current folder is image_20241111_000033.png.
-
 ```
 
 ```npcsh
@@ -785,7 +773,7 @@ and then the associated image :
 An important facet that makes `npcsh` so powerful is the ability to pipe outputs from one tool call to another. This allows for the chaining of commands and the creation of complex workflows. For example, you can use the output of a search to generate an image, or you can use the output of an image analysis to generate a report. Here is an example of how this might look in practice:
 ```npcsh
 npcsh> what is the gdp of russia in 2024? | /vixynt 'generate an image that contains {0}'
-
+```
 ### Executing Bash Commands
 You can execute bash commands directly within npcsh. The LLM can also generate and execute bash commands based on your natural language requests.
 For example:
@@ -1083,7 +1071,10 @@ Search can be accomplished through the `/search` macro. You can specify the prov
 you must set a perplexity api key as an environment variable as described above. The default provider is duckduckgo.
 
 NOTE: while google is an available search engine, they recently implemented changes (early 2025) that make the python google search package no longer as reliable.
-For now, we will use duckduckgo and revisit this issue when other more critical aspects are handled.
+Duckduckgo's search toool also givies rate limit errors often, so until a more robust
+solution is implemented for it, Perplexity's will be the most reliable.
+
+
 
 
 ```npcsh

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/cli.py

@@ -33,6 +33,7 @@ from .llm_funcs import (
     get_stream,
     get_conversation,
 )
+from .plonk import plonk, action_space
 from .search import search_web
 from .shell_helpers import *
 import os
@@ -54,10 +55,12 @@ def main():
         "assemble",
         "build",
         "compile",
+        "chat",
         "init",
         "new",
         "plonk",
         "sample",
+        "search",
         "select",
         "serve",
         "spool",
@@ -165,6 +168,10 @@ def main():
         "directory", nargs="?", default=".", help="Directory to build project in"
     )
 
+    # chat
+    chat_parser = subparsers.add_parser("chat", help="chat with an NPC")
+    chat_parser.add_argument("-n", "--npc_name", help="name of npc")
+
     # Compile command
     compile_parser = subparsers.add_parser("compile", help="Compile an NPC")
     compile_parser.add_argument("path", help="Path to NPC file")
@@ -300,7 +307,7 @@ def main():
 
     # Web search
     search_parser = subparsers.add_parser("search", help="search the web")
-    search_parser.add_argument("query", help="search query")
+    search_parser.add_argument("--query", "-q", help="search query")
     search_parser.add_argument(
         "--search_provider",
         "-sp",
@@ -317,7 +324,7 @@ def main():
 
     # Voice chat
     whisper_parser = subparsers.add_parser("whisper", help="start voice chat")
-    whisper_parser.add_argument("npc_name", help="name of the NPC to chat with")
+    whisper_parser.add_argument("-n", "--npc_name", help="name of the NPC to chat with")
 
     args = parser.parse_args()
 
@@ -367,6 +374,13 @@ def main():
             port=args.port if args.port else 5337,
             cors_origins=cors_origins,
         )
+    elif args.command == "chat":
+        npc_name = args.npc_name
+        npc_path = get_npc_path(npc_name, NPCSH_DB_PATH)
+        current_npc = load_npc_from_file(npc_path, sqlite3.connect(NPCSH_DB_PATH))
+        return enter_spool_mode(
+            model=args.model, provider=args.provider, npc=current_npc
+        )
 
     elif args.command == "init":
         if args.templates:
@@ -395,16 +409,18 @@ def main():
         )
 
     elif args.command == "compile":
-        compile_npc(args.path)
+        npc_compiler = NPCCompiler(npc_directory, NPCSH_DB_PATH)
+        compiled = npc_compiler.compile(args.path)
+        print("NPC compiled to:", compiled)
 
     elif args.command == "plonk":
         task = args.task or args.spell
         npc_name = args.name
-        run_plonk_task(
-            task=task,
-            npc_name=npc_name,
-            model=args.model or NPCSH_REASONING_MODEL,
-            provider=args.provider or NPCSH_REASONING_PROVIDER,
+        plonk(
+            task,
+            action_space,
+            model=args.model or NPCSH_CHAT_MODEL,
+            provider=args.provider or NPCSH_CHAT_PROVIDER,
         )
 
     elif args.command == "sample":
@@ -443,10 +459,14 @@ def main():
             model=args.model,
             provider=args.provider,
         )
-        print(result)
+        print(result["output"])
 
     elif args.command == "whisper":
-        start_whisper_chat(args.npc_name)
+        npc_name = args.npc_name
+        npc_path = get_npc_path(npc_name, NPCSH_DB_PATH)
+        current_npc = load_npc_from_file(npc_path, sqlite3.connect(NPCSH_DB_PATH))
+
+        enter_whisper_mode(npc=current_npc)
 
     elif args.command == "tool":
         result = invoke_tool(

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/image.py

@@ -283,8 +283,6 @@ def analyze_image(
                         api_key=api_key,
                     )
 
-                    print(response)
-                    # Add to command history *inside* the try block
                     return response
 
             except Exception as e:

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/llm_funcs.py

@@ -802,15 +802,20 @@ ReAct choices then will enter reasoning flow
 
     prompt = f"""
     A user submitted this query: {command}
+
     Determine the nature of the user's request:
-    1. Is it a specific request for a task that could be accomplished via a bash command or a simple python script that could be executed in a single bash call?
-    2. Should a tool be invoked to fulfill the request?
-    3. Is it a general question that requires an informative answer or a highly specific question that
+
+    1. Should a tool be invoked to fulfill the request?
+
+    2. Is it a general question that requires an informative answer or a highly specific question that
         requires inforrmation on the web?
-    4. Would this question be best answered by an alternative NPC?
-    5. Is it a complex request that actually requires more than one
-    tool to be called, perhaps in a sequence?
-    6. is there a need for the user to provide additional input to fulfill the request?
+
+    3. Would this question be best answered by an alternative NPC?
+
+    4. Is it a complex request that actually requires more than one
+        tool to be called, perhaps in a sequence?
+
+    5. is there a need for the user to provide additional input to fulfill the request?
 
 
 
@@ -877,8 +882,12 @@ ReAct choices then will enter reasoning flow
 
     prompt += f"""
     In considering how to answer this, consider:
-    - Whether it can be answered via a bash command on the user's computer. e.g. if a user is curious about file sizes within a directory or about processes running on their computer, these are likely best handled by a bash command.
-    - Whether more context from the user is required to adequately answer the question. e.g. if a user asks for a joke about their favorite city but they don't include the city , it would be helpful to ask for that information. Similarly, if a user asks to open a browser and to check the weather in a city, it would be helpful to ask for the city and which website or source to use.
+
+    - Whether more context from the user is required to adequately answer the question.
+        e.g. if a user asks for a joke about their favorite city but they don't include the city ,
+        it would be helpful to ask for that information. Similarly, if a user asks to open a browser
+        and to check the weather in a city, it would be helpful to ask for the city and which website
+        or source to use.
     - Whether a tool should be used.
 
 
@@ -887,14 +896,17 @@ ReAct choices then will enter reasoning flow
     extra tools or agent passes.
     Only use tools or pass to other NPCs
     when it is obvious that the answer needs to be as up-to-date as possible. For example,
-    a question about where mount everest is does not necessarily need to be answered by a tool call or an agent pass.
+        a question about where mount everest is does not necessarily need to be answered by a tool call or an agent pass.
     Similarly, if a user asks to explain the plot of the aeneid, this can be answered without a tool call or agent pass.
-    If a user were to ask for the current weather in tokyo or the current price of bitcoin or who the mayor of a city is, then a tool call or agent pass may be appropriate. If a user asks about the process using the most ram or the biggest file in a directory, a bash command will be most appropriate.
+
+    If a user were to ask for the current weather in tokyo or the current price of bitcoin or who the mayor of a city is,
+        then a tool call or agent pass may be appropriate.
+
     Tools are valuable but their use should be limited and purposeful to
     ensure the best user experience.
 
     Respond with a JSON object containing:
-    - "action": one of ["execute_command", "invoke_tool", "answer_question", "pass_to_npc", "execute_sequence", "request_input"]
+    - "action": one of ["invoke_tool", "answer_question", "pass_to_npc", "execute_sequence", "request_input"]
     - "tool_name": : if action is "invoke_tool": the name of the tool to use.
                      else if action is "execute_sequence", a list of tool names to use.
     - "explanation": a brief explanation of why you chose this action.
@@ -907,7 +919,7 @@ ReAct choices then will enter reasoning flow
 
     The format of the JSON object is:
     {{
-        "action": "execute_command" | "invoke_tool" | "answer_question" | "pass_to_npc" | "execute_sequence" | "request_input",
+        "action": "invoke_tool" | "answer_question" | "pass_to_npc" | "execute_sequence" | "request_input",
         "tool_name": "<tool_name(s)_if_applicable>",
         "explanation": "<your_explanation>",
         "npc_name": "<npc_name(s)_if_applicable>"
@@ -915,7 +927,9 @@ ReAct choices then will enter reasoning flow
 
     If you execute a sequence, ensure that you have a specified NPC for each tool use.
 
-    Remember, do not include ANY ADDITIONAL MARKDOWN FORMATTING. There should be no prefix 'json'. Start straight with the opening curly brace.
+    Remember, do not include ANY ADDITIONAL MARKDOWN FORMATTING.
+    There should be no leading ```json.
+
     """
 
     if docs_context:
@@ -932,11 +946,6 @@ ReAct choices then will enter reasoning flow
         {context}
 
         """
-
-    # print(prompt)
-
-    # For action determination, we don't need to pass the conversation messages to avoid confusion
-    # print(npc, model, provider)
     action_response = get_llm_response(
         prompt,
         model=model,
@@ -965,12 +974,11 @@ ReAct choices then will enter reasoning flow
     else:
         response_content_parsed = response_content
 
-    # Proceed according to the action specified
     action = response_content_parsed.get("action")
     explanation = response_content["explanation"]
-    # Include the user's command in the conversation messages
     print(f"action chosen: {action}")
     print(f"explanation given: {explanation}")
+
     if response_content_parsed.get("tool_name"):
         print(f"tool name: {response_content_parsed.get('tool_name')}")
 
@@ -1316,8 +1324,9 @@ def handle_tool_call(
             stream=stream,
             messages=messages,
         )
-        if "Error" in tool_output:
-            raise Exception(tool_output)
+        if not stream:
+            if "Error" in tool_output:
+                raise Exception(tool_output)
     except Exception as e:
         # diagnose_problem = get_llm_response(
         ##    f"""a problem has occurred.

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/npc_compiler.py

@@ -501,6 +501,18 @@ class SilentUndefined(Undefined):
         return ""
 
 
+class Context:
+    def __init__(self, context=None, mcp_servers=None, databases=None, files=None):
+        self.context = context
+        self.mcp_servers = mcp_servers
+        self.databases = databases
+        self.files = files
+
+    def load_context_file(self, path):
+        with open(path, "r") as f:
+            self.context = yaml.safe_load(f)
+
+
 class Tool:
     def __init__(self, tool_data: dict):
         if not tool_data or not isinstance(tool_data, dict):
@@ -551,6 +563,7 @@ class Tool:
 
         # Process Steps
         for i, step in enumerate(self.steps):
+
             context = self.execute_step(
                 step,
                 context,
@@ -564,6 +577,7 @@ class Tool:
             # if i is the last step and the user has reuqested a streaming output
             # then we should return the stream
             if i == len(self.steps) - 1 and stream:  # this was causing the big issue X:
+                print("tool successful, passing output to stream")
                 return context
         # Return the final output
         if context.get("output") is not None:
@@ -592,8 +606,14 @@ class Tool:
         except Exception as e:
             print(f"Error rendering template: {e}")
             rendered_code = code
-
-        if engine == "natural":
+        # render engine if necessary
+        try:
+            template = jinja_env.from_string(engine)
+            rendered_engine = template.render(**context)
+        except:
+            print("error rendering engine")
+            rendered_engine = engine
+        if rendered_engine == "natural":
             if len(rendered_code.strip()) > 0:
                 # print(f"Executing natural language step: {rendered_code}")
                 if stream:
@@ -610,7 +630,7 @@ class Tool:
                     context["llm_response"] = response_text
                     context["results"] = response_text
 
-        elif engine == "python":
+        elif rendered_engine == "python":
             exec_globals = {
                 "__builtins__": __builtins__,
                 "npc": npc,
@@ -639,12 +659,23 @@ class Tool:
             exec_env = context.copy()
             try:
                 exec(rendered_code, exec_globals, new_locals)
+                exec_env.update(new_locals)
+
+                context.update(exec_env)
+
                 exec_env.update(new_locals)
                 context.update(exec_env)
-                # If output is set, also set it as results
+
+                # Add this line to explicitly copy the output
+                if "output" in new_locals:
+                    context["output"] = new_locals["output"]
+
+                # Then your existing code
                 if "output" in exec_env:
                     if exec_env["output"] is not None:
                         context["results"] = exec_env["output"]
+                        print("result from code execution: ", exec_env["output"])
+
             except NameError as e:
                 tb_lines = traceback.format_exc().splitlines()
                 limited_tb = (

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/npc_team/npcsh.ctx

@@ -1,5 +1,11 @@
-text: |
+context: |
     The npcsh NPC team is devoted to providing a safe and helpful
     environment for users where they can work and be as successful as possible.
     npcsh is a command-line tool that makes it easy for users to harness
-    the power of LLMs from a command line shell.
+    the power of LLMs from a command line shell.
+databases:
+    - ~/npcsh_history.db
+mcp_servers:
+    - /path/to/mcp/server.py
+    - @npm for server
+

npcsh-0.3.31/npcsh/npc_team/tools/bash_executer.tool

@@ -0,0 +1,32 @@
+tool_name: bash_executor
+description: Execute bash queries.
+inputs:
+  - bash_command
+  - user_request
+steps:
+  - engine: python
+    code: |
+      import subprocess
+      import os
+      cmd = '{{bash_command}}'  # Properly quote the command input
+      def run_command(cmd):
+        process = subprocess.Popen(cmd, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        stdout, stderr = process.communicate()
+        if stderr:
+          print(f"Error: {stderr.decode('utf-8')}")
+          return stderr
+        return stdout
+      result = run_command(cmd)
+      output = result.decode('utf-8')
+
+  - engine: natural
+    code: |
+
+      Here is the result of the bash command:
+      ```
+      {{ output }}
+      ```
+      This was the original user request: {{ user_request }}
+
+      Please provide a response accordingly.
+

npcsh-0.3.31/npcsh/npc_team/tools/code_executor.tool

@@ -0,0 +1,16 @@
+tool_name: code_executor
+description: Execute scripts with a specified language. choose from python, bash, R, or javascript. Set the ultimate result as the "output" variable. It must be a string. Do not add unnecessary print statements.
+inputs:
+  - code
+  - language
+steps:
+  - engine: '{{ language }}'
+    code: |
+      {{code}}
+  - engine: natural
+    code: |
+      Here is the result of the code execution that an agent ran.
+      ```
+      {{ output }}
+      ```
+      please provide a response accordingly.

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/npc_team/tools/generic_search.tool

@@ -2,13 +2,21 @@ tool_name: "internet_search"
 description: Searches the web for information based on a query in order to verify timiely details (e.g. current events) or to corroborate information in uncertain situations. Should be mainly only used when users specifically request a search, otherwise an LLMs basic knowledge should be sufficient.
 inputs:
   - query
+  - provider: ''
 steps:
   - engine: "python"
     code: |
       from npcsh.search import search_web
+      from npcsh.npc_sysenv import NPCSH_SEARCH_PROVIDER
       query = "{{ query }}"
+      provider = '{{ provider }}'
+      if provider.strip() != '':
+          results = search_web(query, num_results=5, provider = provider)
+      else:
+          results = search_web(query, num_results=5, provider = NPCSH_SEARCH_PROVIDER)
+
       print('QUERY in tool', query)
-      results = search_web(query, num_results=5)
+      results = search_web(query, num_results=5, provider = NPCSH_SEARCH_PROVIDER)
       print('RESULTS in tool', results)
   - engine: "natural"
     code: |

npcsh-0.3.31/npcsh/npc_team/tools/npcsh_executor.tool

@@ -0,0 +1,9 @@
+tool_name: npcsh_executor
+description: Execute npcsh commands. Use the macro commands.
+inputs:
+  - code
+  - language
+steps:
+  - engine: "{{language}}"
+    code: |
+      {{code}}

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/npc_team/tools/sql_executor.tool

@@ -1,5 +1,5 @@
-tool_name: sql_executor
-description: Execute SQL queries on the ~/npcsh_history.db and display the result. The database contains only information about conversations and other user-provided data. It does not store any information about individual files.
+tool_name: data_pull
+description: Execute queries on the ~/npcsh_history.db to pull data. The database contains only information about conversations and other user-provided data. It does not store any information about individual files.
 inputs:
   - sql_query
   - interpret: false  # Note that this is not a boolean, but a string

{npcsh-0.3.29 → npcsh-0.3.31}/npcsh/search.py

@@ -5,6 +5,7 @@ import os
 
 from bs4 import BeautifulSoup
 from duckduckgo_search import DDGS
+from duckduckgo_search.exceptions import DuckDuckGoSearchException
 
 try:
     from googlesearch import search
@@ -41,7 +42,6 @@ def search_perplexity(
         "max_tokens": max_tokens,
         "temperature": temperature,
         "top_p": top_p,
-        "search_domain_filter": ["perplexity.ai"],
         "return_images": False,
         "return_related_questions": False,
         "search_recency_filter": "month",
@@ -58,6 +58,7 @@ def search_perplexity(
     # Make the POST request to the API
     response = requests.post(url, json=payload, headers=headers)
     response = json.loads(response.text)
+    print(response)
     return [response["choices"][0]["message"]["content"], response["citations"]]
 
 
@@ -89,13 +90,19 @@ def search_web(
 
     if provider == "duckduckgo":
         ddgs = DDGS()
-        search_results = ddgs.text(query, max_results=num_results)
-        print(search_results, type(search_results))
-        urls = [r["href"] for r in search_results]
-        results = [
-            {"title": r["title"], "link": r["href"], "content": r["body"]}
-            for r in search_results
-        ]
+        try:
+            search_results = ddgs.text(query, max_results=num_results)
+            print(search_results, type(search_results))
+            urls = [r["href"] for r in search_results]
+            results = [
+                {"title": r["title"], "link": r["href"], "content": r["body"]}
+                for r in search_results
+            ]
+        except DuckDuckGoSearchException as e:
+            print("DuckDuckGo search failed: ", e)
+            urls = []
+            results = []
+
     else:  # google
         urls = list(search(query, num_results=num_results))
         # google shit doesnt seem to be working anymore, apparently a lbock they made on browsers without js?