[jpvajda]

PR Summary

Fix: String|String[] parameter URL encoding for WebSocket clients

Issue: #629

Problem:
When passing list values to keyterm, keywords, replace, and search parameters in WebSocket clients, the SDK incorrectly URL-encoded the entire list as a string (e.g., keyterm=%5B%27TERM1%27%2C+%27TERM2%27%5D) instead of creating multiple query parameters (e.g., keyterm=TERM1&keyterm=TERM2).

Solution:

• Updated type signatures to accept typing.Union[str, typing.Sequence[str]]
• Added runtime logic to check parameter type and iterate over sequences
• Each list item now correctly adds a separate query parameter

Files Changed:

• src/deepgram/listen/v1/client.py - Fixed: keyterm, keywords, replace, search
• src/deepgram/listen/v1/raw_client.py - Fixed: keyterm, keywords, replace, search
• src/deepgram/listen/v2/client.py - Fixed: keyterm
• src/deepgram/listen/v2/raw_client.py - Fixed: keyterm

Testing:

• ✅ All 467 unit tests pass
• ✅ Validated with reproduction scripts showing correct URL encoding
• ✅ Backward compatible (single strings still work)

Note: REST/media API clients already handled this correctly; only WebSocket clients were affected.

Reproduction Script used for testing

"""
Simple validation test - captures actual WebSocket URLs and validates encoding
"""
import os
import sys
from unittest.mock import patch, MagicMock
from deepgram import DeepgramClient

# Get API key
api_key = os.getenv("DEEPGRAM_API_KEY", "test_key")
client = DeepgramClient(api_key=api_key)

print("="*80)
print("SDK FIX VALIDATION - URL ENCODING TEST")
print("="*80)

results = {}

def test_parameter(test_name, connect_func, expected_url_pattern):
    """Test a parameter and validate the URL encoding"""
    print(f"\n{test_name}")
    print("-"*80)

    captured_url = None

    # Mock the websocket to capture the URL
    with patch('websockets.sync.client.connect') as mock_connect:
        mock_protocol = MagicMock()
        mock_connect.return_value.__enter__ = lambda self: mock_protocol
        mock_connect.return_value.__exit__ = lambda self, *args: None

        try:
            connect_func()
            # Get the URL that was passed to websocket connect
            captured_url = mock_connect.call_args[0][0] if mock_connect.called else None
        except Exception as e:
            print(f"Error during connection: {e}")

    if not captured_url:
        print(f"❌ FAIL - No URL captured")
        results[test_name] = False
        return

    print(f"Captured URL: {captured_url}")
    print(f"Expected pattern: {expected_url_pattern}")

    # Check if the URL contains the expected pattern
    if expected_url_pattern in captured_url:
        print(f"✅ PASS - URL encoding is correct")
        results[test_name] = True
    else:
        print(f"❌ FAIL - URL encoding is incorrect")
        results[test_name] = False

# TEST 1: v1 keyterm with list
def test1():
    with client.listen.v1.connect(model="nova-3", keyterm=["TERM1", "TERM2"]) as conn:
        pass

test_parameter(
    "TEST 1: v1 keyterm=['TERM1', 'TERM2']",
    test1,
    "keyterm=TERM1&keyterm=TERM2"
)

# TEST 2: v1 keywords with list
def test2():
    with client.listen.v1.connect(model="nova-2", keywords=["word1", "word2"]) as conn:
        pass

test_parameter(
    "TEST 2: v1 keywords=['word1', 'word2']",
    test2,
    "keywords=word1&keywords=word2"
)

# TEST 3: v1 replace with list
def test3():
    with client.listen.v1.connect(model="nova-3", replace=["old1:new1", "old2:new2"]) as conn:
        pass

test_parameter(
    "TEST 3: v1 replace=['old1:new1', 'old2:new2']",
    test3,
    "replace=old1%3Anew1&replace=old2%3Anew2"
)

# TEST 4: v1 search with list
def test4():
    with client.listen.v1.connect(model="nova-3", search=["term1", "term2"]) as conn:
        pass

test_parameter(
    "TEST 4: v1 search=['term1', 'term2']",
    test4,
    "search=term1&search=term2"
)

# TEST 5: v2 keyterm with list
def test5():
    with client.listen.v2.connect(
        model="flux-general-en",
        encoding="linear16",
        sample_rate="16000",
        keyterm=["TERM1", "TERM2"]
    ) as conn:
        pass

test_parameter(
    "TEST 5: v2 keyterm=['TERM1', 'TERM2']",
    test5,
    "keyterm=TERM1&keyterm=TERM2"
)

# SUMMARY
print("\n" + "="*80)
print("RESULTS")
print("="*80)

all_passed = all(results.values())
for test_name, passed in results.items():
    status = "✅ PASS" if passed else "❌ FAIL"
    print(f"{status} - {test_name}")

print("="*80)
if all_passed:
    print("🎉 ALL TESTS PASSED")
    sys.exit(0)
else:
    print("❌ SOME TESTS FAILED")
    sys.exit(1)

Summary by CodeRabbit

• New Features
  • Enhanced search and listen parameters: keyterm, keywords, replace, and search now accept multiple values in addition to single values, enabling more powerful and flexible query configurations across all API versions.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

---

• To see the specific tasks where the Asana app for GitHub is being used, see below:
  • https://app.asana.com/0/0/1211917378264536

[coderabbitai]

Walkthrough

The changes extend four query parameters (keyterm, keywords, replace, search) in V1 listen clients and keyterm in V2 listen clients to accept either a single string or a sequence of strings. Parameters are expanded into multiple query entries when sequences are provided, maintaining backward compatibility with single-string inputs.

Changes

Cohort / File(s)	Summary
V1 listen clients – type expansion and multi-value handling src/deepgram/listen/v1/client.py, src/deepgram/listen/v1/raw_client.py	Widened keyterm, keywords, replace, and search parameter types from Optional[str] to Optional[Union[str, Sequence[str]]] in both V1Client/RawV1Client and AsyncV1Client/AsyncRawV1Client connect methods. Updated query parameter construction to iterate over sequences and add multiple entries per parameter.
V2 listen clients – keyterm expansion src/deepgram/listen/v2/client.py, src/deepgram/listen/v2/raw_client.py	Widened keyterm parameter type from Optional[str] to Optional[Union[str, Sequence[str]]] in both V2Client/RawV2Client and AsyncV2Client/AsyncRawV2Client connect methods. Updated parameter serialization to handle both single strings and sequences, adding multiple keyterm entries for sequences.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

• All changes follow a consistent, repetitive pattern across four files
• Type signature widening is mechanical and straightforward
• Parameter iteration logic is consistent and non-complex
• No substantial new control flow or error-handling changes

Possibly related PRs

• support multiple keyterms for v2 listen client #595: Modifies V2 listen connect methods to accept multiple keyterms using similar type expansion and query parameter iteration approach.

Suggested reviewers

• lukeocodes

Pre-merge checks and finishing touches

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically describes the main change: fixing URL encoding issues with string or string array parameters, which directly addresses the core problem resolved in the changeset.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix/keyterms

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 0

🧹 Nitpick comments (4)

src/deepgram/listen/v2/client.py (1)

46-47: Multi-value keyterm handling is correct and backward compatible; docstring slightly out of sync

The widened type and isinstance(keyterm, str) / else: for term in keyterm pattern correctly:

• preserves single-string behavior,
• expands sequences into repeated keyterm=... entries, and
• avoids accidentally iterating over characters for plain strings.

Only minor nit: the keyterm docstrings still state typing.Optional[str], which no longer matches the annotated type. If these docstrings are not fully generator-owned, consider updating them (or the generator) to reflect Optional[Union[str, Sequence[str]]].

Also applies to: 102-107, 161-162, 217-222

src/deepgram/listen/v2/raw_client.py (1)

34-35: Raw V2 keyterm multi-value support matches client behavior; consider docstring alignment

The raw sync/async V2 connect methods now mirror the high-level client behavior:

• strings yield a single keyterm query param,
• sequences yield multiple keyterm params, one per item.

This fixes the list-encoding issue while maintaining compatibility for existing string callers. As with the client, the docstrings for keyterm still advertise typing.Optional[str]; if these docs aren’t purely generated, consider updating them (or the generator) to describe the Optional[Union[str, Sequence[str]]] input.

Also applies to: 90-95, 138-139, 194-199

src/deepgram/listen/v1/client.py (1)

56-57: V1 multi-value handling for keyterm/keywords/replace/search looks good; docs lag the types

The updated signatures and query-param logic for these four fields in both V1Client.connect and AsyncV1Client.connect correctly:

• Preserve old behavior for single strings.
• Expand sequences into repeated query params (e.g., search=foo&search=bar).
• Avoid treating strings as sequences of characters via the isinstance(..., str) guard.
• Keep sync/async behavior consistent.

The effective behavior change for empty lists (now emitting no param rather than a stringified list) is an improvement on the previous broken encoding.

Only small suggestion: the parameter docstrings for keyterm, keywords, replace, and search still say typing.Optional[str]; if feasible, consider updating the generator or docs to reflect Optional[Union[str, Sequence[str]]] so users see the multi-value capability in generated documentation.

Also applies to: 66-68, 169-180, 197-210, 283-285, 293-295, 396-407, 424-437

src/deepgram/listen/v1/raw_client.py (1)

37-38: Raw V1 multi-value parameter behavior is consistent and correct; docstrings could be updated

The raw v1 sync and async connect methods now:

• Accept Optional[Union[str, Sequence[str]]] for keyterm, keywords, replace, and search.
• Correctly emit one query param per value when a sequence is provided, while keeping legacy single-string usage intact.
• Mirror the high-level V1 client behavior so there’s no divergence between raw and wrapped APIs.

As elsewhere, the docstrings for these parameters still show typing.Optional[str]. If you control the generator or these docs, it would be a small win to update them to reflect the union type so that SDK users discover the new multi-value support from the documentation as well.

Also applies to: 47-49, 150-161, 178-191, 243-245, 253-255, 356-367, 384-397

📜 Review details

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 78ac7b7 and a2346e4.

📒 Files selected for processing (4)

• src/deepgram/listen/v1/client.py (8 hunks)
• src/deepgram/listen/v1/raw_client.py (8 hunks)
• src/deepgram/listen/v2/client.py (4 hunks)
• src/deepgram/listen/v2/raw_client.py (4 hunks)

🧰 Additional context used

🧠 Learnings (4)

📚 Learning: 2024-07-01T19:17:04.194Z

Learnt from: dvonthenen
Repo: deepgram/deepgram-python-sdk PR: 426
File: deepgram/clients/listen/v1/__init__.py:36-43
Timestamp: 2024-07-01T19:17:04.194Z
Learning: Unused imports in `deepgram/clients/listen/v1/__init__.py` are retained to maintain backward compatibility and should not be flagged for removal in reviews.

Applied to files:

• src/deepgram/listen/v1/client.py
• src/deepgram/listen/v1/raw_client.py

📚 Learning: 2024-10-09T02:19:46.087Z

Learnt from: dvonthenen
Repo: deepgram/deepgram-python-sdk PR: 426
File: deepgram/clients/listen/v1/rest/options.py:12-12
Timestamp: 2024-10-09T02:19:46.087Z
Learning: Unused imports in `deepgram/clients/listen/v1/rest/options.py` are retained to maintain backwards compatibility.

Applied to files:

• src/deepgram/listen/v1/client.py

📚 Learning: 2024-07-01T19:21:39.778Z

Learnt from: dvonthenen
Repo: deepgram/deepgram-python-sdk PR: 426
File: deepgram/clients/listen/v1/websocket/__init__.py:8-8
Timestamp: 2024-07-01T19:21:39.778Z
Learning: Unused imports in `deepgram/clients/listen/v1/websocket/__init__.py` are retained to maintain backward compatibility and should not be flagged for removal in reviews.

Applied to files:

• src/deepgram/listen/v1/raw_client.py

📚 Learning: 2025-06-22T17:02:32.416Z

Learnt from: lukeocodes
Repo: deepgram/deepgram-python-sdk PR: 543
File: tests/unit_test/test_unit_authentication.py:67-87
Timestamp: 2025-06-22T17:02:32.416Z
Learning: In the Deepgram Python SDK, DeepgramClient("arbitrary-string") should be treated as an API key for backward compatibility with existing code patterns.

Applied to files:

• src/deepgram/listen/v2/raw_client.py

[BrianHillis-DG]

Redaction has a similar problem, the API accepts multiple parameters for different types of redaction like redact=pci&redact=numbers

[lukeocodes]

closing, as i will be making this change in the spec