vimlm 0.0.4__py3-none-any.whl → 0.0.6__py3-none-any.whl

vimlm-0.0.6.dist-info/METADATA

@@ -0,0 +1,158 @@
+Metadata-Version: 2.2
+Name: vimlm
+Version: 0.0.6
+Summary: VimLM - LLM-powered Vim assistant
+Home-page: https://github.com/JosefAlbers/vimlm
+Author: Josef Albers
+Author-email: albersj66@gmail.com
+Requires-Python: >=3.12.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: nanollama==0.0.5
+Requires-Dist: mlx_lm_utils==0.0.2
+Requires-Dist: watchfiles==1.0.4
+Dynamic: author
+Dynamic: author-email
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+
+# VimLM - Local LLM-Powered Coding Assistant for Vim
+
+![vimlm](https://raw.githubusercontent.com/JosefAlbers/VimLM/main/assets/captioned_vimlm.gif)
+
+LLM-powered coding companion for Vim, inspired by GitHub Copilot/Cursor. Integrates contextual code understanding, summarization, and AI assistance directly into your Vim workflow.
+
+## Features
+
+- **Model Agnostic** - Use any MLX-compatible model via a configuration file
+- **Vim-Native UX** - Intuitive keybindings and split-window responses
+- **Deep Context** - Understands code context from:
+    - Current file
+    - Visual selections
+    - Referenced files
+    - Project directory structure
+- **Conversational Coding** - Iterative refinement with follow-up queries
+- **Air-Gapped Security** - 100% offline - no APIs, no tracking, no data leaks
+
+## Requirements
+
+- Apple M-series chip
+- Python 3.12.8
+
+## Quick Start
+
+Install:
+
+```zsh
+pip install vimlm
+```
+
+Launch:
+
+```zsh
+vimlm
+```
+
+or
+
+```zsh
+vimlm path/to/your_file
+```
+
+This launches Vim with the LLM in a split window, ready to assist you.
+
+## Basic Usage
+
+1. **From Normal Mode**:
+    - `Ctrl-l`: Adds current line + file to context
+    - Example prompt: "Regex for removing html tags from item.content"
+
+2. **From Visual Mode**:
+    - Select code → `Ctrl-l`: Adds selected block + current file to context
+    - Example prompt: "Convert this to async/await syntax"
+
+3. **Inline Commands**:
+
+!include: Adds specified outside files/folders to context:
+    - `!include` (no path): Current folder
+    - `!include ~/scrap/jph00/hypermedia-applications.summ.md`: Specific folder
+    - `!include ~/wtm/utils.py`: Specific file
+    - Example prompt: "AJAX-ify this app @ ~/scrap/jph00/hypermedia-applications.summ.md"
+
+!deploy: Extract code blocks to files in user specified dir (current dir if none specified).
+
+!continue: Lets the LLM resume the generation from where it had halted due to length limits.
+
+!followup: Continue the thread (equivalent to `Ctrl-j`
+
+4. **Follow-Up**: After initial response:
+    - `Ctrl-j`: Continue thread
+    - Example follow-up: "In Manifest V3"
+
+4. **Code Extraction: Press `Ctrl-p` to choose a code block from the response and insert them into:
+    - The last selected visual block (in Normal mode)
+    - The current selection (in Visual mode)
+    - Example workflow:
+        1. Select a block of code in Visual mode.
+        2. Prompt the LLM with `Ctrl-l` (e.g., "Convert this to async/await syntax").
+        3. Once the response is generated, press `Ctrl-p` to replace the selected block with the extracted code.
+
+### Key Bindings
+
+| Binding    | Mode          | Action                                 |
+|------------|---------------|----------------------------------------|
+| `Ctrl-l`   | Normal/Visual | Send current file + selection to LLM   |
+| `Ctrl-j`   | Normal        | Continue conversation                  |
+| `Ctrl-p`   | Normal/Visual | Replace the selection with generated code |
+| `Esc`      | Prompt        | Cancel input                           |
+
+## Advanced Configuration
+
+VimLM uses a JSON config file with the following configurable parameters:
+
+```json
+{
+  "DEBUG": true,
+  "LLM_MODEL": null,
+  "NUM_TOKEN": 2000,
+  "SEP_CMD": "!",
+  "USE_LEADER": false
+}
+```
+
+### Custom Model Setup
+
+1. **Browse models**: [MLX Community Models on Hugging Face](https://huggingface.co/mlx-community)
+
+2. **Edit config file**:
+
+```json
+{
+  "LLM_MODEL": "mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit"
+}
+```
+3. **Save to**:
+
+```
+~/vimlm/config.json
+```
+
+4. **Restart VimLM**
+
+
+### Custom Keybinding
+
+If you prefer using `<Leader>` in place of `<Ctrl>` for the ViMLM key bindings:
+
+```json
+{
+  "USER_LEADER": true
+}
+```
+
+

vimlm-0.0.6.dist-info/RECORD

@@ -0,0 +1,7 @@
+vimlm.py,sha256=_2EKVmfvM-Ikf48XzaxvC0rcLCxiHO4f6_fCslAqyCo,21181
+vimlm-0.0.6.dist-info/LICENSE,sha256=f1xgK8fAXg_intwnbc9nLkHf7ODPLtgpHs7DetQHOro,11343
+vimlm-0.0.6.dist-info/METADATA,sha256=INHaQDN1Id04li0LNWmlkVCIEmnKi9ABpbhAYFGokYE,4233
+vimlm-0.0.6.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
+vimlm-0.0.6.dist-info/entry_points.txt,sha256=mU5V4MYsuIzCc6YB-Ro-6USSHWN5vHw8UDnTEoq0isw,36
+vimlm-0.0.6.dist-info/top_level.txt,sha256=I8GjqoiP--scYsO3AfLhha-6Ax9ci3IvbWvVbPv8g94,6
+vimlm-0.0.6.dist-info/RECORD,,

vimlm.py

@@ -17,7 +17,6 @@ import subprocess
 import json
 import os
 from watchfiles import awatch
-from nanollama32 import Chat
 import shutil
 import time
 from itertools import accumulate
@@ -25,13 +24,18 @@ import argparse
 import tempfile
 from pathlib import Path
 from string import Template
+import re
 
 DEBUG = True
+LLM_MODEL = None # "mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit"
 NUM_TOKEN = 2000
-SEP_CMD = '!@#$'
+SEP_CMD = '!'
+USE_LEADER = False
+DO_RESET = True
+SHOW_USER = False 
 VIMLM_DIR = os.path.expanduser("~/vimlm")
 WATCH_DIR = os.path.expanduser("~/vimlm/watch_dir")
-CFG_FILE = "cfg.json"
+CFG_FILE = 'cfg.json'
 LOG_FILE = "log.json"
 LTM_FILE = "cache.json"
 OUT_FILE = "response.md"
@@ -41,40 +45,85 @@ LOG_PATH = os.path.join(VIMLM_DIR, LOG_FILE)
 LTM_PATH = os.path.join(VIMLM_DIR, LTM_FILE)
 OUT_PATH = os.path.join(WATCH_DIR, OUT_FILE) 
 
-if os.path.exists(WATCH_DIR):
-    shutil.rmtree(WATCH_DIR)
-os.makedirs(WATCH_DIR)
-
-try:
-    with open(CFG_PATH, "r") as f:
-        cfg = cfg.load(f)
-    DEBUG = config.get("DEBUG", DEBUG)
-    NUM_TOKEN = config.get("NUM_TOKEN", NUM_TOKEN)
-    SEP_CMD = config.get("SEP_CMD", SEP_CMD)
-except:
-    with open(CFG_PATH, 'w') as f:
-        json.dump(dict(DEBUG=DEBUG, NUM_TOKEN=NUM_TOKEN, SEP_CMD=SEP_CMD), f, indent=2)
-
-def toout(s, key='tovim'):
-    with open(OUT_PATH, 'w', encoding='utf-8') as f:
+def toout(s, key=None, mode=None):
+    key = '' if key is None else ':'+key
+    mode = 'w' if mode is None else mode
+    with open(OUT_PATH, mode, encoding='utf-8') as f:
         f.write(s)
-    tolog(s, key)
+    tolog(s, key='tovim'+key+':'+mode)
 
 def tolog(log, key='debug'):
     if not DEBUG and key == 'debug':
         return
-    if os.path.exists(LOG_PATH):
+    try:
         with open(LOG_PATH, "r", encoding="utf-8") as log_f:
             logs = json.load(log_f)
-    else:
+    except:
         logs = []
     logs.append(dict(key=key, log=log, timestamp=time.ctime()))
     with open(LOG_PATH, "w", encoding="utf-8") as log_f:
         json.dump(logs, log_f, indent=2)
 
+if os.path.exists(WATCH_DIR):
+    shutil.rmtree(WATCH_DIR)
+os.makedirs(WATCH_DIR)
+
+try:
+    with open(CFG_PATH, "r") as f:
+        config = json.load(f)
+    DEBUG = config.get("DEBUG", DEBUG)
+    LLM_MODEL = config.get("LLM_MODEL", LLM_MODEL)
+    NUM_TOKEN = config.get("NUM_TOKEN", NUM_TOKEN)
+    SEP_CMD = config.get("SEP_CMD", SEP_CMD)
+    USE_LEADER = config.get("USE_LEADER", USE_LEADER)
+except Exception as e:
+    tolog(str(e))
+    with open(CFG_PATH, 'w') as f:
+        json.dump(dict(DEBUG=DEBUG, LLM_MODEL=LLM_MODEL, NUM_TOKEN=NUM_TOKEN, SEP_CMD=SEP_CMD, USE_LEADER=USE_LEADER), f, indent=2)
+
 toout('Loading LLM...')
-chat = Chat(variant='uncn_llama_32_3b_it')
-toout('LLM is ready')
+if LLM_MODEL is None:
+    from nanollama import Chat
+    chat = Chat(model_path='uncn_llama_32_3b_it')
+    toout(f'LLM is ready')
+else:
+    from mlx_lm_utils import Chat
+    chat = Chat(model_path=LLM_MODEL)
+    toout(f'{model_path.split('/')[-1]} is ready')
+
+def deploy(dest=None, src=None, reformat=True):
+    prompt_deploy = 'Reformat the text to ensure each code block is preceded by a filename in **filename.ext** format, with only alphanumeric characters, dots, underscores, or hyphens in the filename. Remove any extraneous characters from filenames.'
+    tolog(f'deploy {dest=} {src=} {reformat=}')
+    if src:
+        chat.reset()
+        with open(src, 'r') as f:
+            prompt_deploy = f.read().strip() + '\n\n---\n\n' + prompt_deploy
+    if reformat:
+        response = chat(prompt_deploy, max_new=NUM_TOKEN, verbose=False, stream=False)['text']
+        toout(response, 'deploy')
+        lines = response.splitlines()
+    else:
+        with open(OUT_PATH, 'r') as f:
+            lines = f.readlines()
+    dest = get_path(dest)
+    os.makedirs(dest, exist_ok=True)
+    current_filename = None
+    code_block = []
+    in_code_block = False
+    for line in lines:
+        line = line.rstrip()
+        if line.startswith("```"):
+            if in_code_block and current_filename:
+                with open(os.path.join(dest, os.path.basename(current_filename)), "w", encoding="utf-8") as code_file:
+                    code_file.write("\n".join(code_block) + "\n")
+                code_block = []
+            in_code_block = not in_code_block
+        elif in_code_block:
+            code_block.append(line)
+        else:
+            match = re.match(r"^\*\*(.+?)\*\*$", line)
+            if match:
+                current_filename = re.sub(r"[^a-zA-Z0-9_.-]", "", match.group(1))
 
 def is_binary(file_path):
     try:
@@ -115,7 +164,7 @@ def split_str(doc, max_len=2000, get_len=len):
     return chunks
 
 def retrieve(src_path, max_len=2000, get_len=len):
-    src_path = os.path.expanduser(src_path)
+    src_path = get_path(src_path)
     result = {}
     if not os.path.exists(src_path):
         tolog(f"The path {src_path} does not exist.", 'retrieve')
@@ -126,26 +175,31 @@ def retrieve(src_path, max_len=2000, get_len=len):
                 content = f.read()
             result = {src_path:dict(timestamp=os.path.getmtime(src_path), list_str=split_str(content, max_len=max_len, get_len=get_len))}
         except Exception as e:
-            tolog(f'Skipped {filename} due to {e}', 'retrieve')
-        return result
-    for filename in os.listdir(src_path):
-        try:
-            file_path = os.path.join(src_path, filename)
-            if filename.startswith('.') or is_binary(file_path):
+            tolog(f'Failed to retrieve({filename}) due to {e}')
+    else:
+        for filename in os.listdir(src_path):
+            try:
+                file_path = os.path.join(src_path, filename)
+                if filename.startswith('.') or is_binary(file_path):
+                    continue
+                if os.path.isfile(file_path):
+                    with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:
+                        content = f.read()
+                    result[file_path] = dict(timestamp=os.path.getmtime(file_path), list_str=split_str(content, max_len=max_len, get_len=get_len))
+            except Exception as e:
+                tolog(f'Failed to retrieve({filename}) due to {e}')
                 continue
-            if os.path.isfile(file_path):
-                with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:
-                    content = f.read()
-                result[file_path] = dict(timestamp=os.path.getmtime(file_path), list_str=split_str(content, max_len=max_len, get_len=get_len))
-        except Exception as e:
-            tolog(f'Skipped {filename} due to {e}', 'retrieve')
-            continue
     return result
 
-def get_ntok(s):
-    return len(chat.tokenizer.encode(s)[0])
+def get_path(s):
+    if not s:
+        s = '.'
+    s = s.strip()
+    s = os.path.expanduser(s)
+    s = os.path.abspath(s)
+    return s
 
-def ingest(src):
+def ingest(src, max_len=NUM_TOKEN):
     def load_cache(cache_path=LTM_PATH):
         if os.path.exists(cache_path):
             with open(cache_path, 'r', encoding='utf-8') as f:
@@ -158,53 +212,135 @@ def ingest(src):
                 current_data[k] = v
         with open(cache_path, 'w', encoding='utf-8') as f:
             json.dump(current_data, f, indent=2)
-    toout('Ingesting...')
-    format_ingest = '{volat}{incoming}\n\n---\n\nPlease provide a succint bullet point summary for above:'
-    format_volat = 'Here is a summary of part 1 of **{k}**:\n\n---\n\n{newsum}\n\n---\n\nHere is the next part:\n\n---\n\n'
-    dict_doc = retrieve(src, get_len=get_ntok)
+    src = get_path(src)
+    tolog(f'ingest {src=}')
+    result = ''
+    src_base = os.path.basename(src)
+    if os.path.isdir(src):
+        listdir = [i for i in os.listdir(src) if not i.startswith('.') and '.' in i]
+        result = '\n- '.join([f'--- {src_base} ---', *listdir]) + '\n\n'
+    elif os.path.isfile(src):
+        result = ''
+    else:
+        tolog(f'Failed to ingest({src})')
+        return ''
+    dict_doc = retrieve(src, max_len=max_len, get_len=chat.get_ntok)
+    toout(f'Ingesting {src}...')
+    format_ingest = '{volat}{incoming}\n\n---\n\nPlease provide a succint bullet point summary for above:' 
+    format_volat = 'Here is a summary of part 1 of **{k}**:\n\n---\n\n{newsum}\n\n---\n\nHere is the next part:\n\n---\n\n' 
     dict_sum = {}
     cache = load_cache()
+    max_new_accum = int(max_len/len(dict_doc)) if len(dict_doc) > 0 else max_len
     for k, v in dict_doc.items():
-        v_stamp = v['timestamp']
-        if v_stamp == cache.get(k, {}).get('timestamp'):
-            dict_sum[k] = cache[k]
-            continue
         list_str = v['list_str']
+        v_stamp = v['timestamp']
         if len(list_str) == 0:
             continue
-        if len(list_str) > 1:
-            max_new_sum = int(NUM_TOKEN / len(list_str) / 2)
+        if len(list_str) == 1 and chat.get_ntok(list_str[0]) <= max_new_accum:
+            chat_summary = list_str[0]
+        else:
+            k_base = os.path.basename(k)
+            if v_stamp == cache.get(k, {}).get('timestamp'):
+                dict_sum[k] = cache[k]
+                continue
+            max_new_sum = int(max_len/len(list_str))
             volat = f'**{k}**:\n'
             accum = ''
-            for s in list_str:
+            for i, s in enumerate(list_str):
                 chat.reset()
-                newsum = chat(format_ingest.format(volat=volat, incoming=s.strip()), max_new=max_new_sum, verbose=False, stream=None)[0][:-10].strip()
+                toout(f'\n\nIngesting {k_base} {i+1}/{len(list_str)}...\n\n', mode='a')
+                tolog(f'{s=}') # DEBUG
+                newsum = chat(format_ingest.format(volat=volat, incoming=s.rstrip()), max_new=max_new_sum, verbose=False, stream=OUT_PATH)['text'].rstrip()
+                tolog(f'{k} {i+1}/{len(list_str)}: {newsum=}') # DEBUG
                 accum += newsum + ' ...\n'
                 volat = format_volat.format(k=k, newsum=newsum)
-        else:
-            accum = list_str[0]
-        chat.reset()
-        toout('')
-        chat_summary = chat(format_ingest.format(volat=f'**{k}**:\n', incoming=accum), max_new=int(NUM_TOKEN/4), verbose=False, stream=OUT_PATH)[0][:-10].strip()
-        dict_sum[k] = dict(timestamp=v_stamp, summary=chat_summary)
+            toout(f'\n\nIngesting {k_base}...\n\n', mode='a')
+            tolog(f'{accum=}') # DEBUG
+            if chat.get_ntok(accum) <= max_new_accum:
+                chat_summary = accum.strip()
+            else:
+                chat.reset()
+                chat_summary = chat(format_ingest.format(volat=f'**{k}**:\n', incoming=accum), max_new=max_new_accum, verbose=False, stream=OUT_PATH)['text'].strip()
+        dict_sum[k] = dict(timestamp=v_stamp, summary=chat_summary, ntok=chat.get_ntok(chat_summary))
     dump_cache(dict_sum)
-    result = ''
-    for (k,v) in dict_sum.items():
-        result += f'--- Summary of **{k}** ---\n{v["summary"]}\n\n'
+    for k, v in dict_sum.items():
+        result += f'--- **{os.path.basename(k)}** ---\n{v["summary"].strip()}\n\n'
     result += '---\n\n'
+    toout(result, 'ingest')
     return result
 
-def process_command(data, str_template):
-    try:
-        if len(data['user_command'].strip()) > 0:
-            src = data['user_command']
-        else:
-            src = data['dir']
-        return ingest(src) + str_template
-    except Exception as e:
-        tolog(f'Failed to ingest due to {e}', 'process_command')
-        return str_template
-   
+def process_command(data):
+    if len(data['user']) == 0:
+        response = chat.resume(max_new=NUM_TOKEN, verbose=False, stream=OUT_PATH)
+        toout(response['text'], mode='a')
+        tolog(response)
+        data['user_prompt'] = ''
+        return data
+    if SEP_CMD in data['user']:
+        data['user_prompt'], *cmds = (x.strip() for x in data['user'].split(SEP_CMD))
+    else:
+        data['user_prompt'] = data['user'].strip()
+        cmds = []
+    tolog(f'process_command i {cmds=} {data=}')
+
+    do_reset = False if 'followup' in data else DO_RESET 
+    for cmd in cmds:
+        if cmd.startswith('continue'):
+            arg = cmd.removeprefix('continue').strip('(').strip(')').strip().strip('"').strip("'").strip()
+            data['max_new'] = NUM_TOKEN if len(arg) == 0 else int(arg)
+            response = chat.resume(max_new=data['max_new'], verbose=False, stream=OUT_PATH)
+            toout(response['text'], mode='a')
+            tolog(response)
+            do_reset = False
+            break
+
+        if cmd.startswith('reset'):
+            do_reset = True
+            break
+        if cmd.startswith('followup'):
+            do_reset = False
+            break
+    if do_reset:
+        chat.reset()
+
+    full_path = data['tree']
+    data['dir'] = os.path.dirname(full_path)
+    data['file'] = os.path.basename(full_path)
+    data['ext'] = os.path.splitext(full_path)[1][1:]
+    if chat.stop:
+        data['file'] = ''
+        data['context'] = ''
+    if data['tree'] == OUT_PATH:
+        data['dir'] = os.getcwd()
+        data['file'] = ''
+        data['context'] = ''
+        data['ext'] = ''
+    if data['file'] == '.tmp':
+        data['file'] = ''
+        data['ext'] = ''
+
+    if len(cmds) == 1 and len(cmds[0]) == 0:
+        data['include'] = ingest(data['dir'])
+        return data
+
+    data['include'] = ''
+    for cmd in cmds:
+        if cmd.startswith('include'):
+            arg = cmd.removeprefix('include').strip('(').strip(')').strip().strip('"').strip("'").strip()
+            src = data['dir'] if len(arg) == 0 else arg
+            data['include'] += ingest(src)
+
+    for cmd in cmds:
+        if cmd.startswith('deploy'):
+            arg = cmd.removeprefix('deploy').strip('(').strip(')').strip().strip('"').strip("'").strip()
+            if len(data['user_prompt']) == 0:
+                deploy(dest=arg)
+                data['user_prompt'] = ''
+                return data
+            data['user_prompt'] += "\n\nEnsure that each code block is preceded by a filename in **filename.ext** format. The filename should only contain alphanumeric characters, dots, underscores, or hyphens. Ensure that any extraneous characters are removed from the filenames."
+            data['deploy_dest'] = arg
+    return data
+    
 async def monitor_directory():
     async for changes in awatch(WATCH_DIR):
         found_files = {os.path.basename(f) for _, f in changes}
@@ -220,48 +356,61 @@ async def monitor_directory():
             if 'followup' in os.listdir(WATCH_DIR):
                 os.remove(os.path.join(WATCH_DIR, 'followup'))
                 data['followup'] = True
-                str_template = '{user}'
+            if 'quit' in os.listdir(WATCH_DIR):
+                os.remove(os.path.join(WATCH_DIR, 'quit'))
+                data['quit'] = True
             await process_files(data)
 
 async def process_files(data):
-    tolog(f'{data=}')
-    if 'followup' in data:
-        str_template = "{user}"
-    else:
-        full_path = data['tree']
-        data['dir'] = os.path.dirname(full_path)
-        data['file'] = os.path.basename(full_path)
-        data['ext'] = os.path.splitext(full_path)[1][1:]
-        str_template = ''
-        if len(data['file']) > 0 and data['file'] != '.tmp':
-            str_template += '**{file}**\n'
-        if len(data['context']) > 0:
-            str_template += '```{ext}\n{context}\n```\n\n'
-        if len(data['yank']) > 0:
-            if '\n' in data['yank']:
-                str_template += "```{ext}\n{yank}\n```\n\n"
+    tolog(f'process_files i {data=}')
+    str_template = '{include}'
+    data = process_command(data)
+    if len(data['user_prompt']) == 0:
+        return    
+    if len(data['file']) > 0:
+        str_template += '**{file}**\n'
+    if len(data['context']) > 0 and data['yank'] != data['context']:
+        str_template += '```{ext}\n{context}\n```\n\n'
+    if len(data['yank']) > 0:
+        if '\n' in data['yank']:
+            str_template += "```{ext}\n{yank}\n```\n\n"
+        else:
+            if data['user'] == 0:
+                str_template += "{yank}"
             else:
                 str_template += "`{yank}` "
-        str_template += '{user_prompt}'
-        if SEP_CMD in data['user']:
-            data['user_prompt'], data['user_command'] = (x.strip() for x in data['user'].split(SEP_CMD))
-            str_template = process_command(data, str_template)
-        else:
-            data['user_prompt'] = data['user']
-        chat.reset()
-    tolog(f'{str_template=}') # DEBUG
+    str_template += '{user_prompt}'
+    tolog(f'process_files o {data=}') # DEBUG
+    tolog(f'process_files {str_template=}') # DEBUG
     prompt = str_template.format(**data)
     tolog(prompt, 'tollm')
     toout('')
-    response = chat(prompt, max_new=NUM_TOKEN - get_ntok(prompt), verbose=False, stream=OUT_PATH)
-    toout(response[0][:-10].strip())
-    tolog(response[-1], 'tps')
+    max_new = data['max_new'] if 'max_new' in data else max(10, NUM_TOKEN - chat.get_ntok(prompt))
+    response = chat(prompt, max_new=max_new, verbose=False, stream=OUT_PATH)
+    if SHOW_USER:
+        toout(response['text'])
+    else:
+        toout(response['text'])
+    tolog(response['benchmark'])
+    if 'deploy_dest' in data:
+        deploy(dest=data['deploy_dest'], reformat=False)
+
+mapl, mapj, mapp = ('<Leader>l', '<Leader>j', '<Leader>p') if USE_LEADER else ('<C-l>', '<C-j>', '<C-p>')
 
 VIMLMSCRIPT = Template(r"""
+let s:register_names = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u'] 
 let s:watched_dir = expand('$WATCH_DIR')
 
 function! Monitor()
-    write
+    if exists('s:monitor_timer')
+        call timer_stop(s:monitor_timer)
+        unlet s:monitor_timer
+    endif
+    let response_path = s:watched_dir . '/response.md'
+    let bufnum = bufnr(response_path)
+    if bufnum != -1
+        execute 'bwipeout ' . bufnum
+    endif
     let response_path = s:watched_dir . '/response.md'
     rightbelow vsplit | execute 'view ' . response_path
     setlocal autoread
@@ -283,41 +432,17 @@ function! CheckForUpdates(timer)
 endfunction
 
 function! s:CustomInput(prompt) abort
-    let input = ''
-    let clean_prompt = a:prompt
-    let aborted = v:false
-    echohl Question
-    echon clean_prompt
-    echohl None
-    while 1
-        let c = getchar()
-        if type(c) == v:t_number
-            if c == 13 " Enter
-                break
-            elseif c == 27 " Esc
-                let aborted = v:true
-                break
-            elseif c == 8 || c == 127 " Backspace
-                if len(input) > 0
-                    let input = input[:-2]
-                    echon "\r" . clean_prompt . input . ' '
-                    execute "normal! \<BS>"
-                endif
-            else
-                let char = nr2char(c)
-                let input .= char
-                echon char
-            endif
-        endif
-    endwhile
-    let input_length = aborted ? 0 : len(input)
-    let clear_length = len(clean_prompt) + input_length
-    echon "\r" . repeat(' ', clear_length) . "\r"
-    return aborted ? v:null : input
+    call inputsave()
+    let input = input(a:prompt)
+    call inputrestore()
+    if empty(input)
+        return v:null
+    endif
+    return input
 endfunction
 
-function! SaveUserInput()
-    let user_input = s:CustomInput('Ask LLM: ')
+function! SaveUserInput(prompt)
+    let user_input = s:CustomInput(a:prompt)
     if user_input is v:null
         echo "Input aborted"
         return
@@ -329,37 +454,105 @@ function! SaveUserInput()
     call writefile([current_file], tree_file, 'w')
 endfunction
 
-function! SaveUserInput()
-    let user_input = s:CustomInput('Ask LLM: ')
-    if user_input is v:null
-        echo "Input aborted"
-        return
-    endif
-    let user_file = s:watched_dir . '/user'
-    call writefile([user_input], user_file, 'w')
-    let current_file = expand('%:p')
-    let tree_file = s:watched_dir . '/tree'
-    call writefile([current_file], tree_file, 'w')
+function! VisualPrompt()
+    silent! execute "normal! \<ESC>"
+    silent execute "'<,'>w! " . s:watched_dir . "/yank"
+    silent execute "w! " . s:watched_dir . "/context"
+    " silent! execute "normal! `<V`>"
+    " silent! execute "normal! \<ESC>"
+    call SaveUserInput('VimLM: ')
+endfunction
+
+function! NormalPrompt()
+    silent! execute "normal! V\<ESC>"
+    silent execute "'<,'>w! " . s:watched_dir . "/yank"
+    silent execute "w! " . s:watched_dir . "/context"
+    " silent! execute "normal! \<ESC>"
+    call SaveUserInput('VimLM: ')
 endfunction
 
-function! FollowUpInput()
+function! FollowUpPrompt()
     call writefile([], s:watched_dir . '/yank', 'w')
     call writefile([], s:watched_dir . '/context', 'w')
     call writefile([], s:watched_dir . '/followup', 'w')
-    call SaveUserInput()
+    call SaveUserInput('... ')
+endfunction
+
+function! ExtractAllCodeBlocks()
+    let filepath = s:watched_dir . '/response.md'
+    if !filereadable(filepath)
+        echoerr "File not found: " . filepath
+        return
+    endif
+    let lines = readfile(filepath)
+    let in_code_block = 0
+    let code_blocks = []
+    let current_block = []
+    for line in lines
+        if line =~ '^```'
+            if in_code_block
+                call add(code_blocks, current_block)
+                let current_block = []
+                let in_code_block = 0
+            else
+                let in_code_block = 1
+            endif
+        elseif in_code_block
+            call add(current_block, line)
+        endif
+    endfor
+    if in_code_block
+        call add(code_blocks, current_block)
+    endif
+    for idx in range(len(code_blocks))
+        if idx >= len(s:register_names)
+            break
+        endif
+        let code_block_text = join(code_blocks[idx], "\n")
+        let register_name = s:register_names[idx]
+        call setreg(register_name, code_block_text, 'v') 
+    endfor
+    return len(code_blocks)
 endfunction
 
-vnoremap <C-l> :w! $WATCH_DIR/yank<CR>:w! $WATCH_DIR/context<CR>:call SaveUserInput()<CR>
-nnoremap <C-l> V:w! $WATCH_DIR/yank<CR>:w! $WATCH_DIR/context<CR>:call SaveUserInput()<CR>
-nnoremap <C-r> :call FollowUpInput()<CR>
+function! PasteIntoLastVisualSelection()
+    let num_blocks = ExtractAllCodeBlocks()
+    echo "Extracted " . num_blocks . " blocks into registers @a-@" . s:register_names[num_blocks - 1] . ". Enter register name: "
+    let register_name = nr2char(getchar())
+    if register_name !~ '^[a-z]$'
+        echoerr "Invalid register name. Please enter a single lowercase letter (e.g., a, b, c)."
+        return
+    endif
+    let register_content = getreg(register_name)
+    if register_content == ''
+        echoerr "Register @" . register_name . " is empty."
+        return
+    endif
+    let current_mode = mode()
+    if current_mode == 'v' || current_mode == 'V' || current_mode == ''
+        execute 'normal! "' . register_name . 'p'
+    else
+        normal! gv
+        execute 'normal! "' . register_name . 'p'
+    endif
+endfunction
 
+nnoremap $mapp :call PasteIntoLastVisualSelection()<CR>
+vnoremap $mapp <Cmd>:call PasteIntoLastVisualSelection()<CR>
+vnoremap $mapl <Cmd>:call VisualPrompt()<CR>
+nnoremap $mapl :call NormalPrompt()<CR>
+nnoremap $mapj :call FollowUpPrompt()<CR>
 call Monitor()
-""").safe_substitute(dict(WATCH_DIR=WATCH_DIR))
+""").safe_substitute(dict(WATCH_DIR=WATCH_DIR, mapl=mapl, mapj=mapj, mapp=mapp))
 
 async def main():
     parser = argparse.ArgumentParser(description="VimLM - LLM-powered Vim assistant")
+    parser.add_argument('--test', action='store_true')
     parser.add_argument("vim_args", nargs=argparse.REMAINDER, help="Vim arguments")
     args = parser.parse_args()
+    if args.test:
+        test()
+        return
     with tempfile.NamedTemporaryFile(mode='w', suffix='.vim', delete=False) as f:
         f.write(VIMLMSCRIPT)
         vim_script = f.name
@@ -368,14 +561,17 @@ async def main():
         vim_command.extend(args.vim_args)
     else:
         vim_command.append('.tmp')
-    monitor_task = asyncio.create_task(monitor_directory())
-    vim_process = await asyncio.create_subprocess_exec(*vim_command)
-    await vim_process.wait()
-    monitor_task.cancel()
     try:
-        await monitor_task
-    except asyncio.CancelledError:
-        pass
+        monitor_task = asyncio.create_task(monitor_directory())
+        vim_process = await asyncio.create_subprocess_exec(*vim_command)
+        await vim_process.wait()
+    finally:
+        monitor_task.cancel()
+        try:
+            await monitor_task
+        except asyncio.CancelledError:
+            pass
+        os.remove(vim_script)
 
 def run():
     asyncio.run(main())

vimlm-0.0.4.dist-info/METADATA

@@ -1,82 +0,0 @@
-Metadata-Version: 2.2
-Name: vimlm
-Version: 0.0.4
-Summary: VimLM - LLM-powered Vim assistant
-Home-page: https://github.com/JosefAlbers/vimlm
-Author: Josef Albers
-Author-email: albersj66@gmail.com
-Requires-Python: >=3.12.8
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: nanollama==0.0.3b0
-Requires-Dist: watchfiles==1.0.4
-Dynamic: author
-Dynamic: author-email
-Dynamic: description
-Dynamic: description-content-type
-Dynamic: home-page
-Dynamic: requires-dist
-Dynamic: requires-python
-Dynamic: summary
-
-# VimLM - Vim Language Model Assistant for privacy-conscious developers
-
-![vimlm](https://github.com/user-attachments/assets/67a97048-48df-40c1-9109-53c759e85d96)
-
-An LLM-powered coding companion for Vim, inspired by GitHub Copilot/Cursor. Integrates contextual code understanding, summarization, and AI assistance directly into your Vim workflow.
-
-## Features
-
-- **Real-Time Interaction with local LLMs**: Runs **fully offline** with local models (default: uncensored Llama-3.2-3B).
-- **Integration with Vim's Native Workflow**: Simple Vim keybindings for quick access and split-window interface for non-intrusive responses.
-- **Context-Awareness Beyond Single Files**: Inline support for external documents and project files for richer, more informed responses.
-- **Conversational AI Assistance**: Goes beyond simple code suggestions-explains reasoning, provides alternative solutions, and allows interactive follow-ups. 
-- **Versatile Use Cases**: Not just for coding-use it for documentation lookup, general Q&A, or even casual (uncensored) conversations.
-
-## Installation
-
-```zsh
-pip install vimlm
-```
-
-## Usage
-
-1. Start Vim with VimLM:
-
-```zsh
-vimlm
-```
-
-or
-
-```zsh
-vimlm your_file.js
-```
-
-2. **From Normal Mode**:
-    - `Ctrl-l`: Send current line + file context
-    - Example prompt: "Regex for removing html tags in item.content"
-
-3. **From Visual Mode**:
-    - Select text → `Ctrl-l`: Send selection + file context
-    - Example prompt: "Convert this to async/await syntax"
-
-4. **Add Context**: Use `!@#$` to include additional files/folders:
-    - `!@#$` (no path): Current folder
-    - `!@#$ ~/scrap/jph00/hypermedia-applications.summ.md`: Specific folder
-    - `!@#$ ~/wtm/utils.py`: Specific file
-    - Example prompt: "AJAX-ify this app !@#$ ~/scrap/jph00/hypermedia-applications.summ.md"
-
-5. **Follow-Up**: After initial response:
-    - `Ctrl-r`: Continue thread
-    - Example follow-up: "In Manifest V3"
-
-## Key Bindings
-
-| Binding    | Mode          | Action                                 |
-|------------|---------------|----------------------------------------|
-| `Ctrl-l`   | Normal/Visual | Send current file + selection to LLM   |
-| `Ctrl-r`   | Normal        | Continue conversation                  |
-| `Esc`      | Prompt        | Cancel input                           |
-
-

vimlm-0.0.4.dist-info/RECORD

@@ -1,7 +0,0 @@
-vimlm.py,sha256=kCJphqs6B-ZUGzeikiW3u629Nw5Ih3DlDJV1rsKGXAY,13410
-vimlm-0.0.4.dist-info/LICENSE,sha256=f1xgK8fAXg_intwnbc9nLkHf7ODPLtgpHs7DetQHOro,11343
-vimlm-0.0.4.dist-info/METADATA,sha256=EdZyb9SyLpy0qyK_jILHa7nrd5G8lPxOm6Xfy7NMfiQ,2832
-vimlm-0.0.4.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-vimlm-0.0.4.dist-info/entry_points.txt,sha256=mU5V4MYsuIzCc6YB-Ro-6USSHWN5vHw8UDnTEoq0isw,36
-vimlm-0.0.4.dist-info/top_level.txt,sha256=I8GjqoiP--scYsO3AfLhha-6Ax9ci3IvbWvVbPv8g94,6
-vimlm-0.0.4.dist-info/RECORD,,