-
|
Hi there I followed the instructions on the wiki to integrate VectorCode into code companion. I followed the instructions to modify the opts for code companion like so: opts = {
strategies = {
chat = {
adaptor = "copilot",
tools = {
vectorcode = {
description = "Run VectorCode to retrieve the project context.",
callback = function()
return require("vectorcode.integrations.codecompanion").chat.make_tool()
end,
},
},
slash_commands = {
-- add the vectorcode command here.
codebase = require("vectorcode.integrations").codecompanion.chat.make_slash_command(),
},
},
inline = {
adaptor = "copilot",
},
cmd = {
adaptor = "copilot",
},
},
},
but I got the following error: /Users/beli/.config/nvim/lua/plugins/code_companion.lua:41: module 'vectorcode.integrations' not found:
no field package.preload['vectorcode.integrations']
cache_loader: module vectorcode.integrations not found
cache_loader_lib: module vectorcode.integrations not found
no file './vectorcode/integrations.lua'
no file '/opt/homebrew/share/luajit-2.1/vectorcode/integrations.lua'
no file '/usr/local/share/lua/5.1/vectorcode/integrations.lua'
no file '/usr/local/share/lua/5.1/vectorcode/integrations/init.lua'
no file '/opt/homebrew/share/lua/5.1/vectorcode/integrations.lua'
no file '/opt/homebrew/share/lua/5.1/vectorcode/integrations/init.lua'
no file './vectorcode/integrations.so'
no file '/usr/local/lib/lua/5.1/vectorcode/integrations.so'
no file '/opt/homebrew/lib/lua/5.1/vectorcode/integrations.so'
no file '/usr/local/lib/lua/5.1/loadall.so'
no file './vectorcode.so'
no file '/usr/local/lib/lua/5.1/vectorcode.so'
no file '/opt/homebrew/lib/lua/5.1/vectorcode.so'
no file '/usr/local/lib/lua/5.1/loadall.so'
# stacktrace:
- ~/.config/nvim/lua/plugins/code_companion.lua:41 _in_ **load**
- ~/.config/nvim/lua/config/lazy.lua:17
- .config/nvim/init.lua:2
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 8 replies
-
|
Hi, is your BTW codecompanion released a new extension API which makes this a lot easier. I'm working on migrating to this (the current approach will still work, but with the extension API it'll be easier to set up tools/slash commands). EDIT: opts = {
extensions = {
vectorcode = {
opts = { add_tool = true, add_slash_command = true, tool_opts = {} },
},
}
}And you're good to go! no more manual configuring of the tool and/or slash command! |
Beta Was this translation helpful? Give feedback.
-
|
And is your codecompanion updated to v15+? bc your config looks fine to me |
Beta Was this translation helpful? Give feedback.
-
|
Ahh ok, redownloading VectorCode fixed it for me. Thanks David! That said, could you explain this behavior to me? Why does CodeCompanion asked me to select which project indexed by Vectorcode I would like to use. Shouldn't it be the one that I am in the repo for? |
Beta Was this translation helpful? Give feedback.
-
|
Also when the tool tried to make the query, I got this error. Do I need to start up some server beforehand? Traceback (most recent call last):
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/chromadb/api/base_http_client.py", line 99, in _raise_chroma_error
resp.raise_for_status()
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/httpx/_models.py", line 829, in raise_for_status
raise HTTPStatusError(message, request=request, response=self)
httpx.HTTPStatusError: Server error '500 Internal Server Error' for url 'http://127.0.0.1:53259/api/v2/tenants/default_tenant/databases/default_database/collections/44000398-1e88-4583-bedb-ea948c4fb0a3/query'
For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/500
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/vectorcode/main.py", line 87, in async_main
return_val = await query(final_configs)
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/vectorcode/subcommands/query/__init__.py", line 186, in query
structured_result = await build_query_results(collection, configs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/vectorcode/subcommands/query/__init__.py", line 87, in build_query_results
for identifier in await get_query_result_files(collection, configs):
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/vectorcode/subcommands/query/__init__.py", line 65, in get_query_result_files
results = await collection.query(
^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/chromadb/api/models/AsyncCollection.py", line 215, in query
query_results = await self._client._query(
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/chromadb/telemetry/opentelemetry/__init__.py", line 134, in async_wrapper
return await f(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/chromadb/api/async_fastapi.py", line 585, in _query
resp_json = await self._make_request(
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/chromadb/api/async_fastapi.py", line 132, in _make_request
BaseHTTPClient._raise_chroma_error(response)
File "/home/ir/.local/pipx/venvs/vectorcode/lib/python3.11/site-packages/chromadb/api/base_http_client.py", line 103, in _raise_chroma_error
raise Exception(f"{resp.text} (trace ID: {trace_id})")
Exception: {"error":"RuntimeError('Cannot return the results in a contigious 2D array. Probably ef or M is too small')"} (trace ID: 0) |
Beta Was this translation helpful? Give feedback.
-
|
Hi, this is because I wanted to give the LLM the ability to query an external project (see this discussion for a use case). This means that I can't hardcode the project to query from, and have to rely on prompt engineering to tell the LLM about which project to search from. For example, with the current implementation, I was able to ask the LLM to query from the codecompanion repository while working in the VectorCode repository for suggestions related to tools/slash commands. However, all that stuff is crunched in the system prompt, and LLMs don't always follow all instructions nicely and may lose information. The current project (whichever parent directory that contains extensions = {
vectorcode = {
opts = {
add_tool = true,
tool_opts = {
ls_on_start = true,
},
},
},
}This will add the available collections to the system prompt when this tool initialises, and the model should be able to find the appropriate project from which to query. At the end of the day, most of the behaviours here are achieved by prompt engineering, and it's simply impossible for me to make sure the experience is consistent across different LLMs. |
Beta Was this translation helpful? Give feedback.
-
This is an upstream bug (in chromadb). See this issue: #24. Dropping and re-vectorising usually help. |
Beta Was this translation helpful? Give feedback.
Hi, is your
optstable wrapped in a function like this? If not, this will make therequire('vectorcode.integrations')being loaded before vectorcode is loaded.BTW codecompanion released a new extension API which makes this a lot easier. I'm working on migrating to this (the current approach will still work, but with the extension API it'll be easier to set up tools/slash commands).
EDIT:
The codecompanion extension is live! Now you can simply do this:
And you're good to go! no more manual configuring of the tool and/or slash command!