gptel-curl.arg (gptel-curl--get-args): Increase curl timeout.
Often local LLMs will offload a query to CPU if there is not enough VRAM or in
the case of an unsupported GPU. When a query is offloaded to the CPU responses
can be significantly slower. If curl times out early the user will not get the
response from the LLM back in Emacs.
This change increases the timeout for curl from 60s to 300s to make gptel usable
in slower environments.
Closes#125
README.org: Update README with new information and a multi-llm demo.
gptel.el (gptel-host, gptel--known-backends, gptel--api-key,
gptel--create-prompt, gptel--request-data, gptel--parse-buffer, gptel-request,
gptel--parse-response, gptel--openai, gptel--debug, gptel--restore-state,
gptel, gptel-backend):
Integrate multiple LLMs through the introcution of gptel-backends. Each backend
is composed of two pieces:
1. An instance of a cl-struct, containing connection, authentication and model
information. See the cl-struct `gptel-backend` for details. A separate
cl-struct type is defined for each supported backend (OpenAI, Azure, GPT4All and
Ollama) that inherits from the generic gptel-backend type.
2. cl-generic implementations of specific tasks, like gathering up and
formatting context (previous user queries and LLM responses), parsing responses
or responses streams etc. The four tasks currently specialized this way are
carried out by `gptel--parse-buffer` and `gptel--request-data` (for constructing
the query) and `gptel--parse-response` and `gptel-curl--parse-stream` (for
parsing the response). See their implementations for details. Some effort has
been made to limit the number of times dispatching is done when reading
streaming responses.
When a backend is created, it is registered in the collection
`gptel--known-backends` and can be accessed by name later, such as from the
transient menu.
Only one of these backends is active at any time in a buffer, stored in the
buffer-local variable `gptel-backend`. Most messaging, authentication etc
accounts for the active backend, although there might be some leftovers.
When using `gptel-request` or `gptel-send`, the active backend can be changed or
let-bound.
- Obsolete `gptel-host`
- Fix the rear-sticky property when restoring sessions from files.
- Document some variables (not user options), like `gptel--debug`
gptel-openai.el (gptel-backend, gptel-make-openai, gptel-make-azure,
gptel-make-gpt4all): This file (currently always loaded) sets up the generic
backend struct and includes constructors for creating OpenAI, GPT4All and Azure
backends. They all use the same API so a single set of defgeneric
implemenations suffices for all of them.
gptel-ollama.el (gptel-make-ollama): This file includes the cl-struct,
constructor and requisite defgeneric implementations for Ollama support.
gptel-transient.el (gptel-menu, gptel-provider-variable, gptel--infix-provider,
gptel-suffix-send):
- Provide access to all available LLM backends and models from `gptel-menu`.
- Adjust keybindings in gptel-menu: setting the model and query parameters is
now bound to two char keybinds, while redirecting input and output is bound to
single keys.
gptel.el (gptel--insert-response):
gptel-curl.el (gptel-curl--stream-insert-response): Make the `gptel'
text-property rear-nonsticky so typing after it is recognized as part of the
user prompt.
* gptel.el (gptel-default-mode): Use `fboundp' instead of `featurep' to check if
markdown-mode is available, since the latter requires `markdown-mode' to be
already loaded.
* gptel.el (gptel--system-message, gptel-directives): Try to make
gptel--system-message read from gptel-directives. This doesn't yet work how
we need it to -- changing gptel-directives does not update
gptel--system-message.
* gptel.el (gptel--restore-state): When there is no "GPTEL_BOUNDS"
org property, `read' asks for stdin instead. Fix by only calling
`read' when this property is non-nil.
Thanks to @Elilif for spotting this bug.
* gptel.el (gptel--save-state, gptel--restore-state,
gptel-temperature, gptel-model, gptel-max-tokens,
gptel-directives, gptel--always, gptel--button-buttonize,
gptel--system-message, gptel--bounds): Write gptel parameters as
file-local variables when saving chats in Markdown or text files.
The local variable gptel--bounds stores the locations of the
responses from the LLM. This is not a great solution, but the best
I can think to do without adding more syntax to the document.
Chats can be restored by turning on `gptel-mode'. One of the
problem with this approach is that if the buffer is modified
before `gptel-mode' is turned on, the state data is out of date.
Another problem is that this metadata block as printed in the
buffer can become quite long. A better approach is needed.
Define helper functions `gptel--always' and
`gptel--button-buttonize' to work around Emacs 27.1 support.
* README.org: Mention saving and restoring chats where
appropriate.
* gptel.el (gptel--insert-response, gptel-pre-response-hook): New
user option `gptel-pre-response-hook' that runs before the
response is inserted into the buffer. This can be used to prepare
the buffer in some user-specified way for the response.
* gptel-curl.el (gptel-curl--stream-filter): Run
`gptel-pre-response-hook' before inserting streaming responses.
* gptel-curl.el (gptel-curl-get-response): Don't convert response
into org-mode unless the buffer from which the request originated
is in org-mode. This makes `gptel-default-mode' less binding, and
only used when creating a new chat session with `gptel'. Also,
gptel should now do the right thing depending on whether the
current buffer is in text, Markdown or Org modes.
* gptel.el (gptel-mode, gptel-set-topic, gptel--create-prompt,
gptel-set-topic, gptel--get-topic-start, gptel--get-bounds,
gptel--save-state, gptel--restore-state): Add support for saving
and restoring gptel state for Org buffers. Support for Markdown
buffers is not yet implemented.
`gptel--save-state' and `gptel--restore-state' save and restores
state using Org properties. With `gptel-mode' active, these are
run automatically when saving the buffer or enabling `gptel-mode'
respectively.
The command `gptel-set-topic' can be used to set a topic for the
current heading, which is stored as an Org property. The topic
name is unused (as of now), but the presence of this property
limits the text context sent to ChatGPT to the heading text up to
the cursor position.
Autload `gptel-mode' since the user may want to enable this (to
restore sessions) without having loaded gptel.el.
gptel.el (gptel-request): when using `gptel-request', let-bind
`gptel--system-message' around call to `gptel--create-prompt' when
the prompt argument is null. This allows `gptel-request' to be
used to send the buffer as a prompt with a different system
message from `gptel--system-message' for that buffer.
---------
Co-authored-by: Neil Fulwiler <neil@fulwiler.me>
* gptel.el (gptel-crowdsourced-prompts-file): This file holds
prompts compiled by the community.
* gptel-transient.el (gptel--read-crowdsourced-prompt,
gptel--crowdsourced-prompts, gptel-system-prompt--setup,
gptel--crowdsourced-prompts-url): Fetch crowdsourced system
prompts from https://github.com/f/awesome-chatgpt-prompts and pick
one to use from the transient menu.
* gptel.el (gptel-proxy): Support a proxy when interacting with openai
endpoint. In many organizations the openai api can only be accessed
via proxy. This is easily supported by curl.
gptel-curl.el (gptel-curl--get-args): tidy up `gptel-curl--get-args'.
---------
Co-authored-by: PalaceChan <XXX>
* gptel.el (gptel-directives): Bump required transient version to
0.4.0. Remove placeholder about hard-coded directives from
the docstring for `gptel-directives'.
* gptel-transient.el (gptel-system-prompt,
gptel-system-prompt--setup): Dynamically generate the system
prompt menu in `gptel-menu'. Delete helper functions that
hard-coded the list of system prompts/directives before.
* gptel-transient.el (gptel--system-prompt-programming,
gptel--system-prompt-default, gptel--system-prompt-writing,
gptel--system-prompt-chat, gptel-system-prompt): Define explicit helper
functions to set the system prompt. This is a temporary workaround for #45
until dynamic transients are supported in a new transient release.
* gptel-curl.el (gptel-curl--stream-cleanup): `gptel-post-response-hook' should
run in the buffer that was current when the request was sent. This was not the
case for the curl method (with response streaming). Fixed.
* gptel.el (gptel--insert-response):
* gptel-transient.el (gptel--suffix-send):
* gptel-curl.el (gptel-curl--stream-filter, gptel-curl--stream-insert-response,
gptel-curl--stream-cleanup):
Handle read-only gptel buffers by redirecting the output to a new buffer (that
pops up automatically). To track this,
- the `:position' argument of the INFO plist, which is a marker, is moved to the
new output buffer.
- the `:buffer' argument of the INFO plist is unmodified, it always points to
the buffer that the request originated from.
* gptel-curl.el (gptel-curl-get-response): Set buffer-local model parameters in
the correct (i.e. gptel) buffer, not in Curl's process buffer. This fixes#43.
* gptel-transient.el (gptel--suffix-send, gptel-menu): When
reading input from the minibuffer and sending the output to an
existing gptel session, only use the prompt read from the
minibuffer. Reword the "Overwrite/Delete prompt" option.
* gptel.el (gptel): `gptel' only pops to the gptel buffer when
called interactively.
* gptel-transient.el: Pop up the gptel session buffer when the
prompt is sent to it, but don't select it. Also message the user
that the response has been redirected to the gptel session.
