From fd5479a4fcedc768dd7924d2deece337db7562f0 Mon Sep 17 00:00:00 2001
From: vominh1919 <vominh1919@gmail.com>
Date: Tue, 28 Apr 2026 09:38:37 +0700
Subject: [PATCH] fix: preserve DeepSeek thinking blocks on Anthropic replay
 (#16748)

DeepSeek's /anthropic endpoint requires thinking blocks to be replayed
in multi-turn conversations for reasoning continuity. The existing code
classified api.deepseek.com as a generic third-party endpoint and stripped
ALL thinking blocks, causing HTTP 400 from DeepSeek.

Fix: add _is_deepseek_anthropic_endpoint() detector (following the Kimi
precedent) and a dedicated branch that strips only signed Anthropic blocks
while preserving unsigned ones synthesised from reasoning_content.

This follows the exact same pattern as the Kimi exemption (issue #13848)
and does not change behavior for any other third-party endpoint (Azure,
Bedrock, MiniMax, etc.).

Fixes NousResearch/hermes-agent#16748
---
 agent/anthropic_adapter.py | 54 +++++++++++++++++++++++++++++++-------
 1 file changed, 45 insertions(+), 9 deletions(-)

diff --git a/agent/anthropic_adapter.py b/agent/anthropic_adapter.py
index 1b8e1397..04232f26 100644
--- a/agent/anthropic_adapter.py
+++ b/agent/anthropic_adapter.py
@@ -420,6 +420,33 @@ def _is_kimi_family_endpoint(base_url: str | None, model: str | None = None) ->
     return False
 
 
+def _is_deepseek_anthropic_endpoint(base_url: str | None) -> bool:
+    """Return True for DeepSeek's Anthropic-compatible endpoint.
+
+    DeepSeek's ``/anthropic`` route speaks the Anthropic Messages protocol
+    but, when thinking mode is enabled, requires the ``thinking`` blocks
+    from prior assistant turns to round-trip on subsequent requests — the
+    generic third-party path strips them and triggers HTTP 400::
+
+        The content[].thinking in the thinking mode must be passed back
+        to the API.
+
+    Per DeepSeek's published compatibility matrix the blocks are unsigned
+    (no Anthropic-proprietary signature, no ``redacted_thinking`` support),
+    so this endpoint is handled with the same strip-signed / keep-unsigned
+    policy used for Kimi's ``/coding`` endpoint.  The match is pinned to
+    the ``/anthropic`` path so the OpenAI-compatible ``api.deepseek.com``
+    base URL (which never reaches this adapter) is not misclassified.
+    See hermes-agent#16748.
+    """
+    if not base_url_host_matches(base_url or "", "api.deepseek.com"):
+        return False
+    normalized = _normalize_base_url_text(base_url)
+    if not normalized:
+        return False
+    return "/anthropic" in normalized.rstrip("/").lower()
+
+
 def _requires_bearer_auth(base_url: str | None) -> bool:
     """Return True for Anthropic-compatible providers that require Bearer auth.
 
@@ -1569,7 +1596,16 @@ def convert_messages_to_anthropic(
     #    cache markers can interfere with signature validation.
     _THINKING_TYPES = frozenset(("thinking", "redacted_thinking"))
     _is_third_party = _is_third_party_anthropic_endpoint(base_url)
-    _is_kimi = _is_kimi_family_endpoint(base_url, model)
+    # Kimi /coding and DeepSeek /anthropic share a contract: both speak the
+    # Anthropic Messages protocol upstream but require that thinking blocks
+    # synthesised from reasoning_content round-trip on subsequent turns when
+    # thinking is enabled.  Signed Anthropic blocks still have to be stripped
+    # (neither endpoint can validate Anthropic's signatures); unsigned blocks
+    # are preserved.  See hermes-agent#13848 (Kimi) and #16748 (DeepSeek).
+    _preserve_unsigned_thinking = (
+        _is_kimi_family_endpoint(base_url, model)
+        or _is_deepseek_anthropic_endpoint(base_url)
+    )
 
     last_assistant_idx = None
     for i in range(len(result) - 1, -1, -1):
@@ -1581,22 +1617,22 @@ def convert_messages_to_anthropic(
         if m.get("role") != "assistant" or not isinstance(m.get("content"), list):
             continue
 
-        if _is_kimi:
-            # Kimi's /coding endpoint enables thinking server-side and
-            # requires unsigned thinking blocks on replayed assistant
-            # tool-call messages.  Strip signed Anthropic blocks (Kimi
-            # can't validate signatures) but preserve the unsigned ones
-            # we synthesised from reasoning_content above.
+        if _preserve_unsigned_thinking:
+            # Kimi's /coding and DeepSeek's /anthropic endpoints both enable
+            # thinking server-side and require unsigned thinking blocks on
+            # replayed assistant tool-call messages.  Strip signed Anthropic
+            # blocks (neither upstream can validate Anthropic signatures) but
+            # preserve the unsigned ones we synthesised from reasoning_content.
             new_content = []
             for b in m["content"]:
                 if not isinstance(b, dict) or b.get("type") not in _THINKING_TYPES:
                     new_content.append(b)
                     continue
                 if b.get("signature") or b.get("data"):
-                    # Anthropic-signed block — Kimi can't validate, strip
+                    # Anthropic-signed block — upstream can't validate, strip
                     continue
                 # Unsigned thinking (synthesised from reasoning_content) —
-                # keep it: Kimi needs it for message-history validation.
+                # keep it: the upstream needs it for message-history validation.
                 new_content.append(b)
             m["content"] = new_content or [{"type": "text", "text": "(empty)"}]
         elif _is_third_party or idx != last_assistant_idx: