feat(gateway): pass through upstream finish_reason end-to-end

The gateway hardcoded "stop" on every terminating stream chunk, so
tool-call rounds reported finish_reason=stop and length caps were
invisible to clients. UnifiedChunk now carries finish_reason; adapters
emit it (with empty-string finish reasons like sensenova treated as
non-terminal), standardSSEChunk passes it through for un-adapted
upstreams, [DONE] no longer emits a duplicate reason-less done chunk,
and both streaming paths emit the real reason with "stop" as fallback.

Also vendor sensenova/agentrouter adapters into the repo: they were
WebUI-only uploads and a deploy sync silently removed them while live
AUTO-chain slots still referenced them.

internal/lua/adapters/sensenova.lua

@@ -0,0 +1,108 @@
+-- sensenova API format adapter
+-- streaming: delta only has reasoning_content, no content field
+-- non-streaming: has both content and reasoning_content
+
+local adapter = {}
+adapter.name = "sensenova"
+adapter.version = "1.0.0"
+adapter.endpoint = "/chat/completions"
+adapter.headers = {}
+
+-- Same as openai - strip provider-specific fields
+function adapter.transform_request(raw_body)
+    local ok, req = pcall(json.decode, raw_body)
+    if not ok then return raw_body end
+    req.disable_thinking = nil
+    req.extra_body = nil
+    if req.messages then
+        for _, msg in ipairs(req.messages) do
+            msg.reasoning_content = nil
+        end
+    end
+    return json.encode(req)
+end
+
+-- Same as openai - extract content from response
+function adapter.transform_response(raw_body)
+    local ok, resp = pcall(json.decode, raw_body)
+    if not ok or resp == nil then return raw_body end
+
+    local unified = {
+        content = "",
+        finish_reason = "",
+        token_usage = { prompt = 0, completion = 0, total = 0 }
+    }
+
+    if type(resp.usage) == "table" then
+        unified.token_usage.prompt = resp.usage.prompt_tokens or 0
+        unified.token_usage.completion = resp.usage.completion_tokens or 0
+        unified.token_usage.total = resp.usage.total_tokens or 0
+    end
+
+    if type(resp.choices) == "table" and #resp.choices > 0 then
+        local ch = resp.choices[1]
+        if type(ch.message) == "table" then
+            unified.content = ch.message.content or ""
+            if ch.message.reasoning_content then
+                unified.reasoning_content = ch.message.reasoning_content
+            end
+        end
+        unified.finish_reason = ch.finish_reason or ""
+    end
+
+    return json.encode(unified)
+end
+
+-- Sensenova-specific stream handling
+-- Upstream puts content in reasoning_content only (no content field)
+-- Also sends finish_reason="" (empty string) on every chunk
+function adapter.transform_stream_chunk(raw_chunk)
+    local ok, chunk = pcall(json.decode, raw_chunk)
+    if not ok then return "" end
+
+    local uses = nil
+    if type(chunk.usage) == "table" then
+        uses = {
+            prompt = chunk.usage.prompt_tokens or chunk.usage.prompt or 0,
+            completion = chunk.usage.completion_tokens or chunk.usage.completion or 0,
+            total = chunk.usage.total_tokens or chunk.usage.total or 0,
+        }
+    end
+
+    if not chunk.choices or #chunk.choices == 0 then
+        if uses ~= nil then
+            return json.encode({ usage = uses, done = false })
+        end
+        return ""
+    end
+
+    local delta = chunk.choices[1].delta or {}
+
+    -- Sensenova: delta has reasoning_content but no content
+    local content = delta.content or ""
+    if content == "" and delta.reasoning_content then
+        content = delta.reasoning_content
+    end
+
+    -- Sensenova: finish_reason is "" on every chunk, "stop" on last
+    local fr = chunk.choices[1].finish_reason
+    local done = (fr == "stop" or fr == "length")
+
+    local unified = {
+        content = content,
+        done = done,
+    }
+    if chunk.choices[1].finish_reason and chunk.choices[1].finish_reason ~= "" then
+        unified.finish_reason = chunk.choices[1].finish_reason
+    end
+    if delta.tool_calls then
+        unified.tool_calls = delta.tool_calls
+    end
+    if uses ~= nil then
+        unified.usage = uses
+    end
+
+    return json.encode(unified)
+end
+
+return adapter