fix(gateway): always inject reply-to pointer, not just when quoted text is absent - #13676
Merged
Conversation
…xt is absent The [Replying to: "..."] prefix is disambiguation, not deduplication. When a user explicitly replies to a prior message, the agent needs a pointer to which specific message they're referencing — even when the quoted text already exists somewhere in history. History can contain the same or similar text multiple times; without an explicit pointer the agent has to guess (or answer for both subjects), and the reply signal is silently dropped. Example: in a conversation comparing Japan and Italy, replying to the "Japan is great for culture..." message and asking "What's the best time to go?" — previously the found_in_history check suppressed the prefix because the quoted text was already in history, leaving the agent to guess which destination the user meant. Now the pointer is always present. Drops the found_in_history guard added in #1594. Token overhead is minimal (snippet capped at 500 chars on the new user turn; cached prefix unaffected). Behavior becomes deterministic: reply sent ⇒ pointer present. Thanks to smartyi for flagging this.
alt-glitch
added
type/bug
4 tasks
aj-nt
pushed a commit
to aj-nt/hermes-agent
that referenced
this pull request
May 1, 2026
…xt is absent (NousResearch#13676) The [Replying to: "..."] prefix is disambiguation, not deduplication. When a user explicitly replies to a prior message, the agent needs a pointer to which specific message they're referencing — even when the quoted text already exists somewhere in history. History can contain the same or similar text multiple times; without an explicit pointer the agent has to guess (or answer for both subjects), and the reply signal is silently dropped. Example: in a conversation comparing Japan and Italy, replying to the "Japan is great for culture..." message and asking "What's the best time to go?" — previously the found_in_history check suppressed the prefix because the quoted text was already in history, leaving the agent to guess which destination the user meant. Now the pointer is always present. Drops the found_in_history guard added in NousResearch#1594. Token overhead is minimal (snippet capped at 500 chars on the new user turn; cached prefix unaffected). Behavior becomes deterministic: reply sent ⇒ pointer present. Thanks to smartyi for flagging this.
Luminet2023
pushed a commit
to Luminet2023/hermes-agent
that referenced
this pull request
May 1, 2026
…xt is absent (NousResearch#13676) The [Replying to: "..."] prefix is disambiguation, not deduplication. When a user explicitly replies to a prior message, the agent needs a pointer to which specific message they're referencing — even when the quoted text already exists somewhere in history. History can contain the same or similar text multiple times; without an explicit pointer the agent has to guess (or answer for both subjects), and the reply signal is silently dropped. Example: in a conversation comparing Japan and Italy, replying to the "Japan is great for culture..." message and asking "What's the best time to go?" — previously the found_in_history check suppressed the prefix because the quoted text was already in history, leaving the agent to guess which destination the user meant. Now the pointer is always present. Drops the found_in_history guard added in NousResearch#1594. Token overhead is minimal (snippet capped at 500 chars on the new user turn; cached prefix unaffected). Behavior becomes deterministic: reply sent ⇒ pointer present. Thanks to smartyi for flagging this.
Luminet2023
added a commit
to Luminet2023/hermes-agent
that referenced
this pull request
May 1, 2026
* upstream/main: (160 commits) feat(llm-wiki): port provenance markers, source hashing, and quality signals from llm-wiki-compiler (NousResearch#13700) fix(vision): restore tier-aware Nous vision model selection (NousResearch#13703) fix(vision): route Nous main-provider vision through tier-aware backend test(approval): regression guards for thread-local callback contract fix(security): TUI approval overlay accepts blind keystrokes, CLI thread-local callback invisible to agent chore(release): add Ifkellx to AUTHOR_MAP for PR NousResearch#12687 fix(vision): resolve Nous vision model correctly in auto-detect path docs(delegate): clarify that the parent agent, not the user, populates goal/context (NousResearch#13698) fix(auxiliary): refresh Nous runtime credentials after aux 401s feat(delegate): orchestrator role and configurable spawn depth (default flat) fix(tui): inject VS16 so text-default emoji render as color glyphs feat(gateway): recognize .pdf in MEDIA: tag extraction (NousResearch#13683) test(delegate): make default_toolsets regression test robust to user config docs(delegate): remove default_toolsets from example config and docs refactor(delegate): drop dead default_toolsets from CLI default config feat(image-gen): add GPT Image 2 to FAL catalog (NousResearch#13677) fix(gateway): always inject reply-to pointer, not just when quoted text is absent (NousResearch#13676) fix(skills/llama-cpp): concise description, restore python bindings, fix curl improve llama.cpp skill fix(tui): route skills.manage through the long-handler thread pool ... # Conflicts: # tools/file_tools.py # uv.lock
02356abc
pushed a commit
to 02356abc/hermes-agent
that referenced
this pull request
May 14, 2026
…xt is absent (NousResearch#13676) The [Replying to: "..."] prefix is disambiguation, not deduplication. When a user explicitly replies to a prior message, the agent needs a pointer to which specific message they're referencing — even when the quoted text already exists somewhere in history. History can contain the same or similar text multiple times; without an explicit pointer the agent has to guess (or answer for both subjects), and the reply signal is silently dropped. Example: in a conversation comparing Japan and Italy, replying to the "Japan is great for culture..." message and asking "What's the best time to go?" — previously the found_in_history check suppressed the prefix because the quoted text was already in history, leaving the agent to guess which destination the user meant. Now the pointer is always present. Drops the found_in_history guard added in NousResearch#1594. Token overhead is minimal (snippet capped at 500 chars on the new user turn; cached prefix unaffected). Behavior becomes deterministic: reply sent ⇒ pointer present. Thanks to smartyi for flagging this.
exiao
added a commit
to exiao/hermes-agent
that referenced
this pull request
May 29, 2026
…test - test_reply_snippet_truncated_to_500_chars -> _to_2000_chars (matches new cap) - Fix pre-existing failure: test_no_prefix_when_reply_to_text_is_empty asserted no prefix, but NousResearch#13676 changed behavior to always inject a no-text pointer. Renamed to test_no_text_reply_injects_generic_pointer and updated assertions.
gweeteve
pushed a commit
to gweeteve/hermes-agent
that referenced
this pull request
Jun 2, 2026
…xt is absent (NousResearch#13676) The [Replying to: "..."] prefix is disambiguation, not deduplication. When a user explicitly replies to a prior message, the agent needs a pointer to which specific message they're referencing — even when the quoted text already exists somewhere in history. History can contain the same or similar text multiple times; without an explicit pointer the agent has to guess (or answer for both subjects), and the reply signal is silently dropped. Example: in a conversation comparing Japan and Italy, replying to the "Japan is great for culture..." message and asking "What's the best time to go?" — previously the found_in_history check suppressed the prefix because the quoted text was already in history, leaving the agent to guess which destination the user meant. Now the pointer is always present. Drops the found_in_history guard added in NousResearch#1594. Token overhead is minimal (snippet capped at 500 chars on the new user turn; cached prefix unaffected). Behavior becomes deterministic: reply sent ⇒ pointer present. Thanks to smartyi for flagging this.
waefrebeorn
pushed a commit
to waefrebeorn/slermes
that referenced
this pull request
Jul 2, 2026
…xt is absent (NousResearch#13676) The [Replying to: "..."] prefix is disambiguation, not deduplication. When a user explicitly replies to a prior message, the agent needs a pointer to which specific message they're referencing — even when the quoted text already exists somewhere in history. History can contain the same or similar text multiple times; without an explicit pointer the agent has to guess (or answer for both subjects), and the reply signal is silently dropped. Example: in a conversation comparing Japan and Italy, replying to the "Japan is great for culture..." message and asking "What's the best time to go?" — previously the found_in_history check suppressed the prefix because the quoted text was already in history, leaving the agent to guess which destination the user meant. Now the pointer is always present. Drops the found_in_history guard added in NousResearch#1594. Token overhead is minimal (snippet capped at 500 chars on the new user turn; cached prefix unaffected). Behavior becomes deterministic: reply sent ⇒ pointer present. Thanks to smartyi for flagging this.
This was referenced Aug 5, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The
[Replying to: "..."]prefix is disambiguation, not deduplication — it must be injected whenever the user explicitly replies, even if the quoted text already appears in history.Root cause
_prepare_inbound_message_textingateway/run.pygated the prefix onfound_in_history(introduced in #1594). Intent was to avoid redundancy, but history can contain the same or similar text multiple times, and the prefix's job is to point to which prior message the user is referencing — not to supply the text. When the guard suppresses the prefix, the agent sees a bare user message and has to guess.Concrete failure (thanks @smartYi for flagging):
Changes
gateway/run.py: dropfound_in_historycheck; always prepend[Replying to: "..."]whenreply_to_textandreply_to_message_idare set. Comment explains why.tests/gateway/test_reply_to_injection.py(new): 5 tests covering prefix present when quoted text is absent from history, prefix still present when quoted text IS in history (regression guard), no prefix without reply context, no prefix whenreply_to_textis empty, 500-char snippet truncation.Validation
reply_to_textis Nonescripts/run_tests.sh tests/gateway/test_reply_to_injection.py tests/gateway/test_shared_group_sender_prefix.py— 7 passed.Cost
Token overhead is minimal: snippet capped at 500 chars, applied only to the new user turn, cached prefix unaffected. Behavior becomes deterministic: reply sent ⇒ pointer present.