branch: externals/ellama
commit 42c65d0001746b19fc1fa300e43818c23d6d1669
Merge: e55ed6962d 1d23874ce3
Author: Sergey Kostyaev <[email protected]>
Commit: GitHub <[email protected]>
Merge pull request #355 from
s-kostyaev/add-different-response-process-methods
Add response processing control options
---
README.org | 3 ++
ellama.el | 150 +++++++++++++++++++++++++++++++++++++++++++++++-------------
ellama.info | 58 ++++++++++++-----------
3 files changed, 151 insertions(+), 60 deletions(-)
diff --git a/README.org b/README.org
index e5421e5432..138929a8e9 100644
--- a/README.org
+++ b/README.org
@@ -218,6 +218,9 @@ More sofisticated configuration example:
a corresponding prompt is inserted into a blueprint buffer.
- ~ellama-community-prompts-update-variables~: Prompt user for values of
variables found in current buffer and update them.
+- ~ellama-response-process-method~: Configure how LLM responses are processed.
+ Options include streaming for real-time output, async for asynchronous
+ processing, or skipping every N messages to reduce resource usage.
* Keymap
diff --git a/ellama.el b/ellama.el
index e1af3090d8..7d71fb4183 100644
--- a/ellama.el
+++ b/ellama.el
@@ -225,6 +225,15 @@ PROMPT is a prompt string."
(const :tag "By generating name with reasoning LLM based on prompt."
ellama-generate-name-by-reasoning-llm)
(function :tag "By custom function")))
+(defcustom ellama-response-process-method 'streaming
+ "How to process LLM response.
+If the default streaming method is too resource-heavy, you can try other
+options."
+ :type `(choice
+ (const :tag "Streaming" streaming)
+ (const :tag "Async" async)
+ (integer :tag "Skip every N messages before process one")))
+
(defcustom ellama-define-word-prompt-template "Define %s"
"Prompt template for `ellama-define-word'."
:type 'string)
@@ -1437,39 +1446,114 @@ failure (with BUFFER current).
(require 'spinner)
(spinner-start ellama-spinner-type))
(let* ((handler (ellama--handle-partial insert-text insert-reasoning
reasoning-buffer))
- (request (llm-chat-streaming
- provider
- llm-prompt
- handler
- (lambda (response)
- (let ((text (plist-get response :text))
- (reasoning (plist-get response :reasoning)))
- (funcall handler response)
- (when (or ellama--current-session
- (not reasoning))
- (kill-buffer reasoning-buffer))
- (with-current-buffer buffer
- (accept-change-group ellama--change-group)
- (when ellama-spinner-enabled
- (spinner-stop))
- (if (and (listp donecb)
- (functionp (car donecb)))
- (mapc (lambda (fn) (funcall fn text))
- donecb)
- (funcall donecb text))
- (when ellama-session-hide-org-quotes
- (ellama-collapse-org-quotes))
- (setq ellama--current-request nil)
- (ellama-request-mode -1))))
- (lambda (_ msg)
- (with-current-buffer buffer
- (cancel-change-group ellama--change-group)
- (when ellama-spinner-enabled
- (spinner-stop))
- (funcall errcb msg)
- (setq ellama--current-request nil)
- (ellama-request-mode -1)))
- t)))
+ (request (pcase ellama-response-process-method
+ ('async (llm-chat-async
+ provider
+ llm-prompt
+ (lambda (response)
+ (let ((text (plist-get response :text))
+ (reasoning (plist-get response
:reasoning)))
+ (funcall handler response)
+ (when (or ellama--current-session
+ (not reasoning))
+ (kill-buffer reasoning-buffer))
+ (with-current-buffer buffer
+ (accept-change-group
ellama--change-group)
+ (when ellama-spinner-enabled
+ (spinner-stop))
+ (if (and (listp donecb)
+ (functionp (car donecb)))
+ (mapc (lambda (fn) (funcall fn
text))
+ donecb)
+ (funcall donecb text))
+ (when ellama-session-hide-org-quotes
+ (ellama-collapse-org-quotes))
+ (setq ellama--current-request nil)
+ (ellama-request-mode -1))))
+ (lambda (_ msg)
+ (with-current-buffer buffer
+ (cancel-change-group
ellama--change-group)
+ (when ellama-spinner-enabled
+ (spinner-stop))
+ (funcall errcb msg)
+ (setq ellama--current-request nil)
+ (ellama-request-mode -1)))
+ t))
+ ('streaming (llm-chat-streaming
+ provider
+ llm-prompt
+ handler
+ (lambda (response)
+ (let ((text (plist-get response :text))
+ (reasoning (plist-get response
:reasoning)))
+ (funcall handler response)
+ (when (or ellama--current-session
+ (not reasoning))
+ (kill-buffer reasoning-buffer))
+ (with-current-buffer buffer
+ (accept-change-group
ellama--change-group)
+ (when ellama-spinner-enabled
+ (spinner-stop))
+ (if (and (listp donecb)
+ (functionp (car donecb)))
+ (mapc (lambda (fn) (funcall fn
text))
+ donecb)
+ (funcall donecb text))
+ (when
ellama-session-hide-org-quotes
+ (ellama-collapse-org-quotes))
+ (setq ellama--current-request nil)
+ (ellama-request-mode -1))))
+ (lambda (_ msg)
+ (with-current-buffer buffer
+ (cancel-change-group
ellama--change-group)
+ (when ellama-spinner-enabled
+ (spinner-stop))
+ (funcall errcb msg)
+ (setq ellama--current-request nil)
+ (ellama-request-mode -1)))
+ t))
+ ((pred integerp)
+ (let* ((cnt 0)
+ (skip-handler
+ (lambda (request)
+ (if (= cnt ellama-response-process-method)
+ (progn
+ (funcall handler request)
+ (setq cnt 0))
+ (cl-incf cnt)))))
+ (llm-chat-streaming
+ provider
+ llm-prompt
+ skip-handler
+ (lambda (response)
+ (let ((text (plist-get response :text))
+ (reasoning (plist-get response
:reasoning)))
+ (funcall handler response)
+ (when (or ellama--current-session
+ (not reasoning))
+ (kill-buffer reasoning-buffer))
+ (with-current-buffer buffer
+ (accept-change-group ellama--change-group)
+ (when ellama-spinner-enabled
+ (spinner-stop))
+ (if (and (listp donecb)
+ (functionp (car donecb)))
+ (mapc (lambda (fn) (funcall fn text))
+ donecb)
+ (funcall donecb text))
+ (when ellama-session-hide-org-quotes
+ (ellama-collapse-org-quotes))
+ (setq ellama--current-request nil)
+ (ellama-request-mode -1))))
+ (lambda (_ msg)
+ (with-current-buffer buffer
+ (cancel-change-group ellama--change-group)
+ (when ellama-spinner-enabled
+ (spinner-stop))
+ (funcall errcb msg)
+ (setq ellama--current-request nil)
+ (ellama-request-mode -1)))
+ t))))))
(with-current-buffer buffer
(setq ellama--current-request request)))))))
diff --git a/ellama.info b/ellama.info
index f2e7da9c17..006f9ad305 100644
--- a/ellama.info
+++ b/ellama.info
@@ -320,6 +320,10 @@ File: ellama.info, Node: Commands, Next: Keymap, Prev:
Installation, Up: Top
buffer.
• ‘ellama-community-prompts-update-variables’: Prompt user for values
of variables found in current buffer and update them.
+ • ‘ellama-response-process-method’: Configure how LLM responses are
+ processed. Options include streaming for real-time output, async
+ for asynchronous processing, or skipping every N messages to reduce
+ resource usage.
File: ellama.info, Node: Keymap, Next: Configuration, Prev: Commands, Up:
Top
@@ -1418,33 +1422,33 @@ Tag Table:
Node: Top1379
Node: Installation3613
Node: Commands8621
-Node: Keymap14965
-Node: Configuration17798
-Node: Context Management23397
-Node: Transient Menus for Context Management24305
-Node: Managing the Context25919
-Node: Considerations26694
-Node: Minor modes27287
-Node: ellama-context-header-line-mode29275
-Node: ellama-context-header-line-global-mode30100
-Node: ellama-context-mode-line-mode30820
-Node: ellama-context-mode-line-global-mode31668
-Node: Ellama Session Header Line Mode32372
-Node: Enabling and Disabling32941
-Node: Customization33388
-Node: Ellama Session Mode Line Mode33676
-Node: Enabling and Disabling (1)34261
-Node: Customization (1)34708
-Node: Using Blueprints35002
-Node: Key Components of Ellama Blueprints35621
-Node: Creating and Managing Blueprints36228
-Node: Variable Management37209
-Node: Keymap and Mode37678
-Node: Transient Menus38614
-Node: Running Blueprints programmatically39160
-Node: Acknowledgments39747
-Node: Contributions40460
-Node: GNU Free Documentation License40844
+Node: Keymap15207
+Node: Configuration18040
+Node: Context Management23639
+Node: Transient Menus for Context Management24547
+Node: Managing the Context26161
+Node: Considerations26936
+Node: Minor modes27529
+Node: ellama-context-header-line-mode29517
+Node: ellama-context-header-line-global-mode30342
+Node: ellama-context-mode-line-mode31062
+Node: ellama-context-mode-line-global-mode31910
+Node: Ellama Session Header Line Mode32614
+Node: Enabling and Disabling33183
+Node: Customization33630
+Node: Ellama Session Mode Line Mode33918
+Node: Enabling and Disabling (1)34503
+Node: Customization (1)34950
+Node: Using Blueprints35244
+Node: Key Components of Ellama Blueprints35863
+Node: Creating and Managing Blueprints36470
+Node: Variable Management37451
+Node: Keymap and Mode37920
+Node: Transient Menus38856
+Node: Running Blueprints programmatically39402
+Node: Acknowledgments39989
+Node: Contributions40702
+Node: GNU Free Documentation License41086
End Tag Table
