llm logs --id-gt and --id-gte options, closes #801

simonw · Feb 28, 2025 · 3a60290 · 3a60290
1 parent 48f67f4
commit 3a60290
Show file tree

Hide file tree

Showing 4 changed files with 35 additions and 5 deletions.
diff --git a/docs/help.md b/docs/help.md
@@ -316,6 +316,8 @@ Options:
   --xl, --extract-last        Extract last fenced code block
   -c, --current               Show logs from the current conversation
   --cid, --conversation TEXT  Show logs for this conversation ID
+  --id-gt TEXT                Return responses with ID > this
+  --id-gte TEXT               Return responses with ID >= this
   --json                      Output logs as JSON
   --help                      Show this message and exit.
 ```

diff --git a/docs/logging.md b/docs/logging.md
@@ -85,7 +85,7 @@ Or `-n 0` to see everything that has ever been logged:
 ```bash
 llm logs -n 0
 ```
-You can truncate the display of the prompts and responses using the `-t/--truncate` option. This can help make the JSON output more readable:
+You can truncate the display of the prompts and responses using the `-t/--truncate` option. This can help make the JSON output more readable - though the `--short` option is usually better.
 ```bash
 llm logs -n 1 -t --json
 ```
@@ -179,6 +179,19 @@ llm logs -q 'cheesecake'
 ```
 The most relevant terms will be shown at the bottom of the output.
 
+(logging-filter-id)=
+
+### Filtering past a specific ID
+
+If you want to retrieve all of the logs that were recorded since a specific response ID you can do so using these options:
+
+- `--id-gt $ID` - every record with an ID greater than $ID
+- `--id-gte $ID` - every record with an ID greater than or equal to $ID
+
+IDs are always issued in ascending order by time, so this provides a useful way to see everything that has happened since a particular record.
+
+This can be particularly useful when {ref}`working with schema data <schemas-logs>`, where you might want to access every record that you have created using a specific `--schema` but exclude records you have previously processed.
+
 (logging-filter-model)=
 
 ### Filtering by model

diff --git a/llm/cli.py b/llm/cli.py
@@ -971,6 +971,8 @@ def logs_turn_off():
     "--conversation",
     help="Show logs for this conversation ID",
 )
+@click.option("--id-gt", help="Return responses with ID > this")
+@click.option("--id-gte", help="Return responses with ID >= this")
 @click.option(
     "json_output",
     "--json",
@@ -996,6 +998,8 @@ def logs_list(
     extract_last,
     current_conversation,
     conversation_id,
+    id_gt,
+    id_gte,
     json_output,
 ):
     "Show recent logged prompts and their responses"
@@ -1069,6 +1073,10 @@ def logs_list(
         where_bits.append("responses.model = :model")
     if conversation_id:
         where_bits.append("responses.conversation_id = :conversation_id")
+    if id_gt:
+        where_bits.append("responses.id > :id_gt")
+    if id_gte:
+        where_bits.append("responses.id >= :id_gte")
     schema_id = None
     if schema:
         schema_id = make_schema_id(schema)[0]
@@ -1087,6 +1095,8 @@ def logs_list(
                 "query": query,
                 "conversation_id": conversation_id,
                 "schema_id": schema_id,
+                "id_gt": id_gt,
+                "id_gte": id_gte,
             },
         )
     )
@@ -1209,7 +1219,9 @@ def logs_list(
                 "# {}{}\n{}".format(
                     row["datetime_utc"].split(".")[0],
                     (
-                        "    conversation: {}".format(row["conversation_id"])
+                        "    conversation: {} id: {}".format(
+                            row["conversation_id"], row["id"]
+                        )
                         if should_show_conversation
                         else ""
                     ),

diff --git a/tests/test_llm_logs.py b/tests/test_llm_logs.py
@@ -81,6 +81,7 @@ def schema_log_path(user_path):
 
 
 datetime_re = re.compile(r"\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}")
+id_re = re.compile(r"id: \w+")
 
 
 @pytest.mark.parametrize("usage", (False, True))
@@ -94,9 +95,11 @@ def test_logs_text(log_path, usage):
     output = result.output
     # Replace 2023-08-17T20:53:58 with YYYY-MM-DDTHH:MM:SS
     output = datetime_re.sub("YYYY-MM-DDTHH:MM:SS", output)
+    # Replace id: whatever with id: xxx
+    output = id_re.sub("id: xxx", output)
     expected = (
         (
-            "# YYYY-MM-DDTHH:MM:SS    conversation: abc123\n\n"
+            "# YYYY-MM-DDTHH:MM:SS    conversation: abc123 id: xxx\n\n"
             "Model: **davinci**\n\n"
             "## Prompt\n\n"
             "prompt\n\n"
@@ -107,7 +110,7 @@ def test_logs_text(log_path, usage):
         )
         + ("## Token usage:\n\n2 input, 5 output\n\n" if usage else "")
         + (
-            "# YYYY-MM-DDTHH:MM:SS    conversation: abc123\n\n"
+            "# YYYY-MM-DDTHH:MM:SS    conversation: abc123 id: xxx\n\n"
             "Model: **davinci**\n\n"
             "## Prompt\n\n"
             "prompt\n\n"
@@ -116,7 +119,7 @@ def test_logs_text(log_path, usage):
         )
         + ("## Token usage:\n\n2 input, 5 output\n\n" if usage else "")
         + (
-            "# YYYY-MM-DDTHH:MM:SS    conversation: abc123\n\n"
+            "# YYYY-MM-DDTHH:MM:SS    conversation: abc123 id: xxx\n\n"
             "Model: **davinci**\n\n"
             "## Prompt\n\n"
             "prompt\n\n"