chore: add docker-compose #113
Conversation
Co-authored-by: John Pruitt <[email protected]> Signed-off-by: Iain Cox <[email protected]>
chore: language update for install from source.
Fixed error in how we describe models available via pgai Signed-off-by: Avthar Sewrathan <[email protected]>
Move asdf plpython3u instructions to plpython3u section
* "Internal" Python code is now supported * New versions do not install Python dependencies system-wide * Each version's Python code and dependencies are installed to separate target locations * The path to the Python libs can be overridden via postgres.conf * SQL source code is either idempotent or incremental * Create extension-owned ai schema * Add migration table to manage incremental SQL * Revised make to facilitate all of the above * Lint sql with pgspot
Signed-off-by: Mike Freedman <[email protected]>
Signed-off-by: Mike Freedman <[email protected]>
Add space in README
Fix typo referencing OpenAI instead of Ollama
environement -> environment Signed-off-by: Ikko Eltociear Ashimine <[email protected]>
…in from local 2. merge to main branch
…ke_pizza.sql rm doc/rag_demo_* rm docs/pgai_rag*
refactor: use monorepo structure * refactor: update .gitignore paths * refactor: path fixes and clarifications in build.py * chore: fix formatting in extension tests * refactor: update paths in test_contents * refactor: update paths in test_dump_restore * refactor: update paths in test_privileges * refactor: update paths in test_cli * fix: format sql identifiers in test_vectorizer * refactor: calc vcr cassette dir * refactor: use single tiktoken_cache across tests * fix: handle when tictoken_cache is str in cli * chore: fix formating and reenable ci check * chore: run ci vectorizer tests not in docker --------- Co-authored-by: John Pruitt <[email protected]>
alejandrodnm
requested review from
jgpruitt,
cevian,
avthars and
a team
as code owners
Adds a Docker compose example. Installs the pgai extension in the when building the Docker container. Enables auto prefix for the env vars of the vectorizer cli options.
alejandrodnm
force-pushed
the
adn/docker-compose
branch
from
05fb532
to
0f8ef77
Compare
| COPY . . | ||
| RUN make install |
There was a problem hiding this comment.
this may cause issues with the existing developer experience. Also, if you're going to do this, you have to do RUN make build-install.
There was a problem hiding this comment.
I added it because we are not providing a Dockerfile that once it's built it contains the extension installed. This was the closes thing, it has everything except the extension installed.
One option is to replace this with the timescale-ha image once that contains the extension installed.
Why would you think it would cause issues? The way I was thinking is that we install on container creation, but then uses can just make changes and re-install with the workflow you wrote in the docs.
About the why adding docker-compose. Many projects do it as way to provide a way to run things locally. Granted we can do that with build.py, installing the extension and the vectorizer. This provides a more "real world" scenario, since each component runs separately, instead of in the same container.
Also, a docker-compose file is something most people are familiar with and can easily wrap their minds around it.
There was a problem hiding this comment.
If we want to go with docker compose, that's fine, but we need all the tooling to work with it. We should pick ONE way and support it fully. I don't want multiple ways that work slightly differently and incompletely. For example, make psql-shell does not work with this docker compose setup.
Adds a Docker compose example.
Installs the pgai extension when building the Docker container.
Enables auto prefix for the env vars of the vectorizer cli options.