Debian Codegraph @

TODO

raw
1 Speech Dispatcher TODO
2 ======================
3
4 The release versions are not final, and could change. Targetted release is
5 based on demand from users, and difficulty of the work involved.
6
7 * Add pitch as an option for capitalization presentation
8 see https://github.com/brailcom/speechd/issues/24
9 * Allow setting a synthesis voice in the user config using spd-conf.
10 * Add support for Mimic
11 see https://github.com/brailcom/speechd/issues/19
12 * Drop getting information from other clients
13 see https://github.com/brailcom/speechd/issues/335
14 (0.11) Migrate to GSettings.
15 (0.11) Synthesizer specific settings API.
16 (0.11) Use more GLib in the server.
17 (0.11) Client audio retrieval API.
18 (0.11) Server to module protocol documentation.
19 (0.12) Server to module protocol improvements.
20 * Move synth modules to plugin architecture with plugin host.
21 * Synth plugin API.
22 * Allow for building synth plugins out of tree.
23 (0.11) Integrate with logind/consolekit.
24 (0.11) Properly support system-wide mode.
25 * Support spawning the server via Systemd socket activation.
26
27
28 The above improvements are documented in detail below. If work has started on
29 a particular project, a git branch will be noted. These git branches are
30 located at https://github.com/TheMuso/speechd-wip.git. To read the most up to
31 date copy of this file, please clone the master Speech Dispatcher git
32 repository, located at git://git.freebsoft.org/git/speechd.gitand check out
33 the master branch.
34
35 Migrate to GSettings
36 --------------------
37
38 * Write the GSettings metadata XML file.
39 * Migrate the server to GSettings.
40 * Listen to GSettings changes.
41 * Migrate synthesizer modules to GSettings.
42 * Write a program to migrate user settings to GSettings.
43
44
45 Synthesizer specific settings API
46 Depends on: Migration to GSettings
47 ---------------------------------
48
49 Background:
50 * Currently have API for espeak pitch range in git master, but this is only
51 useful for espeak.
52 * Espeak module has a config option to show variants along with available
53 voices, which can be a very long list and can choak some clients.
54
55 * Implement server to module protocol to support:
56 - Request available settings.
57 - Request available settings and their current value.
58 - Request the value of a setting.
59 - Set a setting.
60 - Reset a setting to its default.
61 * Implement SSIP protocol support.
62 * Implement C API, see synthesizer specific settings C API draft.
63 * Implement python API.
64
65 Synthesizer specific settings C API draft
66
67 typedef struct {
68 char *name;
69 char *description; /* This should be localized */
70 enum SynthSettingValueType get_type;
71 enum SynthSettingValueType set_type;
72 int min_value;
73 int max_value;
74 char **value_list;
75 void *cur_value;
76 ] SynthSetting;
77
78 In the C API, a NULL terminated array of this structure would be returned for
79 all settings a synth offers.
80
81 The SynthSettingValueType enum would look something like this:
82
83 typedef enum {
84 SYNTH_SETTING_VALUE_UNKNOWN = 0,
85 SYNTH_SETTING_VALUE_NUMBER = 1,
86 SYNTH_SETTING_VALUE_STRING = 2,
87 SYNTH_SETTING_VALUE_STRING_LIST = 3 /* A list of strings for the user
88 to choose from, i.e voice variants */
89 } SynthSettingValueType;
90
91 C API methods to work with these data types could be as follows:
92
93 SynthSetting **spd_synth_get_settings(SPDConnection *connection);
94 int spd_synth_set_setting(SPDConnection *connection, SynthSetting *setting,
95 void *value);
96 void free_synth_settings(SynthSettings **settings);
97
98
99 Use more GLib in the server
100 ---------------------------
101
102 * Use GLib event loops where possible.
103 * use GLib GThreads and GAsyncQueues for thread communication.
104 * Use g_spawn calls for executing modules.
105 * Support multiple client connection methods, unix socket, inet socket.
106 * Use g_debug and other relevant GLib logging facilities for
107 messages/logging.
108 * Use GThreadedSocketService for handling client connections.
109 * Replace custom implementations of parsing buffers with GLib equivalent
110 methods where possible.
111
112
113 Move audio into server
114 ----------------------
115
116 * Consider using a separate socket for audio transfer, however this may be
117 difficult when attempting to synchronise with index marks. An alternative is
118 to send index mark data via the audio socket as well.
119 * Rework modules supporting audio output to not use any advanced internal
120 playback queueing, and simply send the audio in relatively small buffers to
121 the server. Smaller buffers to allow the server to stop/pause the audio more
122 responsively.
123 * Extend priority system to be either global priority, or priority per audio
124 output device.
125 * Rework pulseaudio output to use a GLib event loop.
126 * Rework other audio output modules to better work within an event loop.
127
128
129 Client audio retrieval API
130 --------------------------
131
132 * Allow client to either request audio directly, or have audio written to a
133 designated file on disk.
134 * Allow modules to decline the use of direct audio retrieval. I know of one
135 speech synth that is not supported by speech dispatcher, who's licensing
136 model doesn't allow for direct audio retrieval. If this module is ever
137 supported, its code will likely remain closed to prevent people working
138 around the implementation, but it would still be nice to support this synth
139 in the longer term. (Luke Yelavich)
140 * Load a new instance of the requested synth module, and spin up a worker
141 thread to handle audio file writing or sending to client, to allow the server
142 to dispatch other speech messages, as direct audio retrieval should be
143 independant of the priority system.
144
145
146 Server to module protocol documentation
147 ---------------------------------------
148
149 * Similar to the SSIp documentation, write up a texi document that explains
150 the server to module protocol, currently over stdin/stdout, but may use other
151 IPC in the future.
152
153
154 Server to module protocol improvements
155 --------------------------------------
156
157 * Consider using sockets for IPC, with a dedicated socket per module.
158 * Consider implementing shared memory support, particularly for audio data
159 transfer, but this may depend on whether GLib has a shared memory API, The
160 GMappedFile API may be useful, if the initiator can change the contents of
161 the GMappedFile, and the other side can notice changes. Needs investigation.
162 * Support the launching of modules via systems other than Speech Dispatcher,
163 useful where containers of some sort are being used, and the environment
164 requires that any separate processes are run in containers/other kind of
165 sandbox, hense the use of sockets as per above.
166
167
168 Integrate with logind/consolekit
169 (Depends on migration to GSettings, GLib main event loops everywhere)
170 --------------------------------
171
172 * Query current user, and currently running sessions for that user.
173 * Subscribe to tty change events and cork audio playback and synthesis flow
174 if none of the user's sessions are active.
175 * Allow the enabling/disabling of logind/consolekit via GSettings and at
176 runtime, enabled being the default.
177 * Allow the disabling of consolekit/logind at build time.
178 * Consider abstracting this functionality into plugins, or at the very least
179 separate code with an internal API to more easily support any future
180 session/seat monitoring systems.
181
182
183 Properly support system-wide mode
184 ---------------------------------
185
186 * Set a default user and group for the system wide instance to run under, at
187 build time, and runtime.
188 * Add a systemd unit to allow the use of system wide mode, disabled by
189 default.
190
191
192 Support spawning the server via Systemd socket activation
193 ---------------------------------------------------------
194
195 * Allow this to be enabled/disabled at build time.
196
197
198 Copyright (C) 2001 Brailcom, o.p.s
199 Copyright (C) 2016 Luke Yelavich <[email protected]>
200 Copyright (C) 2018-2021 Samuel Thibault <[email protected]>
201
202 This program is free software; you can redistribute it and/or modify it under
203 the terms of the GNU General Public License as published by the Free Software
204 Foundation; either version 2 of the License, or (at your option) any later
205 version.
206
207 This program is distributed in the hope that it will be useful, but WITHOUT ANY
208 WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
209 PARTICULAR PURPOSE. See the GNU General Public License for more details (file
210 COPYING in the root directory).
211
212 You should have received a copy of the GNU General Public License
213 along with this program. If not, see <https://www.gnu.org/licenses/>.