diff --git a/a2a.yaml b/a2a.yaml
index a3b111a..531dcde 100644
--- a/a2a.yaml
+++ b/a2a.yaml
@@ -1,5 +1,5 @@
 name: quote-judge-studio-v1
-version: 0.1.1
+version: 0.1.2
 entrypoint: agent:QuoteJudgeStudioV1
 expose:
   public: false
diff --git a/agent.py b/agent.py
index 81fdf4f..dc2d8fe 100644
--- a/agent.py
+++ b/agent.py
@@ -48,6 +48,7 @@ MAX_QUOTES = 20
 MAX_VENDOR_LENGTH = 120
 MAX_UPLOADS = 5
 MAX_UPLOAD_BYTES = 128_000
+MAX_UPLOAD_BASE64_CHARS = ((MAX_UPLOAD_BYTES + 2) // 3) * 4
 ALLOWED_UPLOAD_MEDIA_TYPES = {"text/csv", "text/plain", "application/json"}
 _DATABASE_NAME = "quote-judge-studio-v1-data"
 _LOCAL_COMPARISONS: dict[tuple[str, str], dict[str, Any]] = {}
@@ -79,20 +80,12 @@ class QuoteWeights(BaseModel):
     delivery: float = Field(..., ge=0, le=100)
     warranty: float = Field(..., ge=0, le=100)
 
-    @field_validator("warranty")
-    @classmethod
-    def weights_must_be_positive(cls, value: float, info: Any) -> float:
-        values = info.data
-        total = float(values.get("price", 0)) + float(values.get("delivery", 0)) + float(value)
-        if total <= 0:
-            raise ValueError("at least one weight must be positive")
-        return value
 
 
 class BrowserQuoteDocument(BaseModel):
     filename: str = Field(..., min_length=1, max_length=160)
     media_type: str = Field(..., min_length=1, max_length=100)
-    data_base64: str = Field(..., min_length=1)
+    data_base64: str = Field(..., min_length=1, max_length=MAX_UPLOAD_BASE64_CHARS)
 
     @field_validator("filename")
     @classmethod
@@ -115,7 +108,7 @@ class QuoteJudgeStudioV1(A2AAgent[QuoteJudgeStudioV1Config, PlatformUserAuth]):
         "QuoteJudge compares vendor quotes with weighted price, delivery, and "
         "warranty scoring, persists per-user comparisons, and serves a one-page product UI."
     )
-    version = "0.1.1"
+    version = "0.1.2"
 
     config_model = QuoteJudgeStudioV1Config
     auth_model = PlatformUserAuth
@@ -152,7 +145,7 @@ class QuoteJudgeStudioV1(A2AAgent[QuoteJudgeStudioV1Config, PlatformUserAuth]):
             "weights; persist the recommendation and a production receipt."
         ),
         timeout_seconds=30,
-        idempotent=True,
+        idempotent=False,
         cost_class="cheap",
     )
     async def compare_quotes(
@@ -171,6 +164,24 @@ class QuoteJudgeStudioV1(A2AAgent[QuoteJudgeStudioV1Config, PlatformUserAuth]):
             "weights": weights.model_dump(),
         }
 
+        if weights.price + weights.delivery + weights.warranty <= 0:
+            result = {
+                "ok": False,
+                "code": "positive_weight_required",
+                "message": "Set at least one weight above zero before comparing.",
+                "comparison_id": cleaned_id,
+            }
+            await _persist_receipt(
+                tenant_key=tenant,
+                comparison_id=cleaned_id,
+                skill_name="compare_quotes",
+                input_payload=input_payload,
+                result_payload=result,
+                status="validation_error",
+                started_at=started,
+            )
+            return result
+
         if len(quotes) < 2:
             result = {
                 "ok": False,
@@ -222,7 +233,7 @@ class QuoteJudgeStudioV1(A2AAgent[QuoteJudgeStudioV1Config, PlatformUserAuth]):
     @a2a.tool(
         description="Reopen a saved QuoteJudge comparison for the authenticated user.",
         timeout_seconds=15,
-        idempotent=True,
+        idempotent=False,
         cost_class="cheap",
     )
     async def get_comparison(
@@ -261,7 +272,7 @@ class QuoteJudgeStudioV1(A2AAgent[QuoteJudgeStudioV1Config, PlatformUserAuth]):
             "compare extracted quotes, and persist the result."
         ),
         timeout_seconds=30,
-        idempotent=True,
+        idempotent=False,
         cost_class="cheap",
     )
     async def compare_uploaded_quotes_browser(
@@ -287,7 +298,7 @@ class QuoteJudgeStudioV1(A2AAgent[QuoteJudgeStudioV1Config, PlatformUserAuth]):
             "file and compare extracted quotes."
         ),
         timeout_seconds=30,
-        idempotent=True,
+        idempotent=False,
         cost_class="cheap",
     )
     async def compare_uploaded_quote_file(
@@ -342,6 +353,8 @@ class QuoteJudgeStudioV1(A2AAgent[QuoteJudgeStudioV1Config, PlatformUserAuth]):
 def _tenant_key(ctx: RunContext[PlatformUserAuth]) -> str:
     auth = ctx.auth
     ident = auth.user_id if auth.user_id is not None else (auth.sub or auth.email)
+    if ident is None:
+        raise ValueError("authenticated user identity is required")
     return f"user:{ident}"
 
 
diff --git a/db/migrations/002_quotejudge_schema_repair.sql b/db/migrations/002_quotejudge_schema_repair.sql
new file mode 100644
index 0000000..580e07a
--- /dev/null
+++ b/db/migrations/002_quotejudge_schema_repair.sql
@@ -0,0 +1,10 @@
+ALTER TABLE quote_comparisons
+    ADD COLUMN IF NOT EXISTS recommended_vendor TEXT,
+    ADD COLUMN IF NOT EXISTS score DOUBLE PRECISION,
+    ADD COLUMN IF NOT EXISTS normalized_total DOUBLE PRECISION;
+
+ALTER TABLE quote_execution_receipts
+    ADD COLUMN IF NOT EXISTS skill_name TEXT,
+    ADD COLUMN IF NOT EXISTS status TEXT,
+    ADD COLUMN IF NOT EXISTS input_hash TEXT,
+    ADD COLUMN IF NOT EXISTS result_hash TEXT;