89 Commits

Author SHA1 Message Date
CJACK
fd0ec29991 refactor: generalize DSML tag parsing to tolerate model noise; split tiktoken by build tags
Replace hardcoded DSML typo variant lists in Go/Node tool call parsers with
generalized prefix consumption that tolerates repeated leading <, repeated DSML
prefix noise, and trailing pipe terminators. Split tiktoken-dependent token
counting into a build-tagged file for non-cgo platform compatibility. Add /data
directory to Dockerfile for bind-mount permissions.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-01 15:17:11 +08:00
CJACK.
63e62fd1b0 Merge pull request #372 from shern-point/feat/accurate-context-token-length
Feat/accurate context token length
2026-04-30 02:11:32 +08:00
shern-point
78fdd63470 feat: add full-context token regression coverage and docs
Lock in the current_input_file regression with API-level tests and document that returned context token counts now track full prompt semantics with conservative sizing.
2026-04-30 00:46:06 +08:00
shern-point
bd41c8a90c feat: add tokenizer-based token counting utilities
Use go-tiktoken with embedded vocabularies for accurate BPE token counting. CountPromptTokens applies conservative padding so returned context token counts stay slightly above the real value instead of undercounting.
2026-04-30 00:44:11 +08:00
MiY
241334c658 Fix stream compatibility and vision model exposure 2026-04-29 20:23:13 +08:00
shern-point
206c3d5479 fix: apply string protection in shared tool formatters
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-04-28 13:27:41 +08:00
CJACK
90ce595325 chore: update project files 2026-04-27 02:09:11 +08:00
CJACK
abc96a37d8 refactor backend API structure 2026-04-26 06:58:20 +08:00
CJACK
7475defeca fix: align tool call protocol and thinking controls 2026-04-26 04:26:51 +08:00
CJACK
f13ad231ac 全局统一映射 2026-04-26 01:58:15 +08:00
CJACK
131ca7d398 feat: revamp DeepSeek v4 model handling
- replace legacy DeepSeek ids with the new deepseek-v4 model family\n- move thinking control to request parameters and preserve assistant reasoning content\n- switch history split to IGNORE transcript injection and map upload auth failures to 401\n- update admin defaults, API docs, samples, and tests for the new model scheme
2026-04-26 00:02:14 +08:00
ouqiting
9a404e75fc feat: 对话记录支持保存并展示 HISTORY 内容 2026-04-23 14:47:43 +08:00
CJACK.
2788e20f05 feat: implement history split functionality to optimize context usage and add corresponding UI settings 2026-04-22 18:23:09 +00:00
CJACK
0f2b5fee23 refactor: enhance XML tool call parsing to support nested structures, CDATA, and repeated tags 2026-04-19 19:58:45 +08:00
CJACK
6688e0ba35 refactor: remove unnecessary whitespace and end-of-sentence markers to align with official DeepSeek chat template encoding 2026-04-19 17:47:45 +08:00
CJACK
0c644d1f4d refactor: remove legacy function call support and simplify tool sieve logic 2026-04-19 04:38:48 +08:00
CJACK
caafdedb00 feat: implement OpenAI-compatible file upload and reference handling for DeepSeek API 2026-04-12 23:30:22 +08:00
CJACK.
433a3a877d feat(prompt): align DeepSeek prompt assembly with tokenizer-style turns 2026-04-12 13:59:42 +08:00
Shuaihao
ba76a2163b Add default, expert, and vision DeepSeek model families 2026-04-08 14:37:22 +08:00
CJACK.
a8c160b05d fix: parse DeepSeek accumulated_token_usage robustly and stabilize lint 2026-04-06 11:14:48 +08:00
CJACK
1530246e4f refactor: move tool call parsing and formatting logic to a dedicated internal/toolcall package 2026-04-06 03:19:18 +08:00
CJACK
2a6b787f38 feat: implement mandatory DeepSeek turn termination markers for system, user, assistant, and tool roles 2026-04-05 21:48:18 +08:00
CJACK
585d35e592 refactor: standardize prompt markers and remove legacy EOS and system instructions tags 2026-04-05 20:50:12 +08:00
CJACK
dc912419c4 Revert "refactor: rename tool XML wrapper from tool_calls to tool_batch and add schema attention blocks to tool prompts"
This reverts commit ade648033d.
2026-04-05 20:08:04 +08:00
CJACK
ade648033d refactor: rename tool XML wrapper from tool_calls to tool_batch and add schema attention blocks to tool prompts 2026-04-05 19:22:43 +08:00
CJACK
b8e9ca2028 refactor: stop stripping _raw and _xml fields from tool call inputs to preserve raw parameter data 2026-04-05 18:33:15 +08:00
CJACK
e37ed38dc8 refactor: remove instruction regarding text-before-tool output in tool prompt documentation 2026-04-05 17:51:26 +08:00
CJACK
5d59775051 refactor: enforce strict XML-only output for tool calls and remove mixed-content instructions 2026-04-05 17:25:52 +08:00
CJACK
22efd8178b refactor: clarify tool call XML placement and add strict rules against result fabrication and internal narration 2026-04-05 17:13:59 +08:00
CJACK
088a750338 feat: add strict schema enforcement and strip hallucinated fields from tool call parameters 2026-04-05 16:59:39 +08:00
CJACK
47dc121690 refactor: improve prompt construction by enforcing explicit newline boundaries between role markers and message content 2026-04-05 04:44:46 +08:00
CJACK.
3497d5d019 fix: classify empty upstream and tighten xml tool-name parsing 2026-04-04 02:14:39 +08:00
CJACK.
0847091864 fix: avoid taking generic <name> as xml tool name 2026-04-04 01:52:57 +08:00
CJACK.
c6340354ec fix: keep execute_command args from xml parameters blocks 2026-04-04 01:42:31 +08:00
CJACK.
07578f9c56 fix tool prompt parameter examples for exec tools 2026-04-02 13:09:41 +08:00
CJACK.
bfca84c2c7 Align tool-call parsing across Go/JS and pass quality gates 2026-04-01 01:24:55 +08:00
TesseractLHY
fe8232bfc1 Fixes bad tool call 2026-03-31 11:16:13 -04:00
CJACK
30a53b6c43 refactor: remove legacy TOOL_CALL_HISTORY/TOOL_RESULT_HISTORY markers and consolidate tool call formatting into a new prompt package 2026-03-30 00:20:38 +08:00
CJACK
621599f8ad test: update message preparation tests to expect explicit User role markers 2026-03-29 19:41:03 +08:00
CJACK
1d6a8e7008 refactor: centralize tool-calling instructions into a shared utility and update Claude/OpenAI adapters to use the unified format. 2026-03-29 16:05:35 +08:00
CJACK.
56a3ed19e8 fix(toolcall): support canonical xml params and guard json shadowing 2026-03-29 11:15:52 +08:00
CJACK.
958f4e39b5 feat(toolcall): prioritize XML for model output and parsing 2026-03-29 10:53:38 +08:00
CJACK.
97a81c4191 Harden toolcall leak interception for function-style payloads 2026-03-22 20:07:12 +08:00
CJACK.
1676c8e4f2 Add backward-compatible aliases for renamed fenced-example tests 2026-03-22 16:25:03 +08:00
CJACK.
b108a7915a Support nested fenced blocks in stream fence tracking 2026-03-22 15:12:55 +08:00
CJACK.
3fccec0e22 test: remove unused asFloat helper 2026-03-22 10:24:11 +08:00
CJACK.
00d38f1187 fix: parse claude tool_use function/parameter format 2026-03-22 09:58:29 +08:00
CJACK.
88a79f212d Fix path control-char repair on JSON fallback parses 2026-03-20 02:52:27 +08:00
CJACK.
acfb3b225d Split toolcall input parsing to satisfy line gate 2026-03-20 02:37:23 +08:00
CJACK.
99a6164000 Fix path corruption when parsing tool call JSON strings 2026-03-20 02:31:37 +08:00